<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="http://subversion.tigris.org">Subversion</a> continues to dig up new tricks from dark recesses up it's sleeve to befuddle the unwary. I ran into another one of these after refactoring some java code.</p>

            <p>Some classes that had been placed in a package named com.raditha.colors were moved to com.raditha.ui since the old name was inappropriate and the old package was deleted. Since then subversion refuses to accept commits on this trunk even though it will happily allow me to carry out commits in other projects. The error message goes something like this:</p>

            <p><code style="color: #000033">Transmitting file data ........svn: Commit failed (details follow):</code></p>

            <p><code style="color: #000033">svn: Invalid change ordering: new node revision ID without delete</code></p>

            <p>&nbsp;</p>

            <p>As usual there is precious little in google, and the often tried remedy of svn cleanup has no effect either. This is the home remedey that I tried out with success to overcome the problem, hope you find it usefull.</p>

            <p>&nbsp;</p>


            <ol>
                <li>Copy the offending folders to another location.
                    <p>&nbsp;</p>

                    <p>&nbsp;</p>
                </li>
                <li>Make a fresh checkout of the offending folder. Lo and behold you will find that the deleted files still linger.
                    <p>&nbsp;</p>

                    <p>&nbsp;</p>
                </li>
                <li>Kill them again. and commit the changes.
                    <p>&nbsp;</p>

                    <p>&nbsp;</p>
                </li>
                <li>Clean up the .svn folders in the files that you backed up in step 1.
                    <p>&nbsp;</p>

                    <p>&nbsp;</p>
                </li>
                <li>Move the remainder into the space that is under version control.
                    <p>&nbsp;</p>

                    <p>&nbsp;</p>
                </li>
                <li>do an svn stat and add the files that are marked as not been under version control.
                    <p>&nbsp;</p>

                    <p>&nbsp;</p>
                </li>
                <li>Do another commit. Viola! the commit is now accepted.
                    <p>&nbsp;</p>

                    <p>&nbsp;</p>
                </li>
            </ol>

            <p>&nbsp;</p>

            <p>&nbsp;</p>


        </div>
    </div>


</body>

</html>

Subversion tricks

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <pre> 07:53:50 up 6 days, 15:32,  1 user,  load average: 0.48, 0.18, 0.06</pre>
            <p></p>
            <p>That result from an inovation for uptime ladies and gentles shows that this computer was runing continously for 6 days and 15 hours without being restarted. Say what? well this is not a server. This is a desktop computer. Don't forget that I install new software on my linux machine on an almost daily basis.</p>
            <p>Would you be able to keep a computer running for even day without a reboot if it was based on windows? it seems whenever you install even the most trivial software, windows screams out to be restarted.</p>
            <p>So why was the machine shut down? I had to go to <a href="http://www.webquarry.com/~raditha/srilanka/travel/galle.html">Galle</a> to look into the well being of one of my aunts and here family effected by the <a href="https://www.raditha.com/~raditha/blog/archives/000586.html">boxing day tsunami</a>.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

uptime

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I have written about the futility of attempting to speed of applications by compiler options or small code level changes. They simply do not work. If you look at the PHP mailing list archives you will find countless thread arguing the merits and demerits of require vs include or <a href="407.html">echo vs print</a> etc.</p>
            <p>These arguments are just a waste of time. In this backdrop I decided to start writing an article with detailed statistics to prove this point. You will find the first part <a href="../../wiki/Readfile_vs_include">here</a>..</p>
            <p>In fact the first step of results were mildly suprising but then the first set of tests are not typical of what you might encounter in a real web application. The second set of tests on the other hand is a better match and the results as expected goes to show that there is little difference between require, include and the other 'file inclusion' functions of PHP.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

PHP require vs include and time wasting

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Many people would dearly love to switch from windows to linux but circumstances prevent them from doing so. Similarly many people would stop using <a href="https://www.mysql.com">mysql</a> and switch to <a href="https://postgresql.org">postgresql</a> if not for the strings that hold them back.</p>
            <p>I fall into this latter category and though I have postgres installed on my computer and use it for lots of things, I found myself upgrading the mysql installation, which had gone quite stale.</p>
            <p>The last few times I installed mysql I used binary downloads but this time around decided to compile from source. As in some earlier versions the documentation suggests that you pass in some optimization parameters to GCC so that you can get better performance out of mysql.</p>
            <p>I have no faith in these compiler options being able to boost the speed of applications but decided to put it to the test just to see if there is any truth in it. The first first time around mysql was compiled with out any optimizations and then I ran a rather <a href="536.html">expensive subquery</a>. Next mysql was compiled with the following compiler options:</p>
            <p><code>-O3 -mtune=pentium4 -march=pentium4</code></p>
            <p>Is there an increase in speed? no. The query takes about 2 minutes regardless of the compiler options.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

mysql upgrade

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Having installed PearPc a couple of days ago, I decided to have a go at installing <a href="http://opendarwin.org">opendarwin</a> as a guest operating system. It turned out to be a much more pleasent experience than <a href="000578.html">installation of opendarwin</a> as a OS that stands on it's own feet.</p>
            <p>PearPC is a PowerPC emulator for linux. You can install opendarwin as well as OS X as a guest operating system for a PearPc host. Qemu is another virtualization system that can emulate several different processors ranging from i386 to ARM. However it has not reached the same levels maturity as pearpc. Qemu is worth wathchig but I digress...</p>
            <p>Installation of opendawin as a client os is well <a href="http://pearpc.sourceforge.net/installdarwin.html">documented.</a> in fact the pearpc documentation is far superior to darwin's own doc. Not having to give darwin direct access to the hardisk means you need less courage to try the installation. After all the known issues page does say it may not work with IDE drives.</p>
            <p>Though the installation takes quite a bit of time when you run it under the emulator it's considerably less painfull than having to constantly reboot the machine. The end result was pleasent as well. darwin did boot up when run as a guest system and I managed to login and get a shell. Now to figure out how xwindow could be started.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

opendarwin on pearpc

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I tried my hand at setting up <a href="http://www.opendarwin.org">opendarwin</a>, the process was a bit scary, the shivers begin when you have a look at the fdisk program that looks so different from what you find on linux. (what command do you use to make a new partition?). I reckoned the safe approach would be to plug in a different hard disk and let the auto partitioning system have it's way with it.</p>
            <p>The installation thereafter proceeded in double quick time. After all there is only one CD. The problem is in the booting. No matter what you do the system would refuse to boot. All you get is a screen of a chewed off applet with a ticker on it that keeps going round and round in circles nothing more. No error messages, no warnings no beeps.</p>
            <p>The kernel supposedly accepts additional parameters such as location of root device etc, unfortunately it does not seem to do accpet them and the documentation (the little that exists) is rather cryptic.</p>
            <p>Finally I came across one doc that casually mentioned that opendarwin may not work with IDE drives. Mmm, now I suppose I will have to get a scsi drive or will this go into my USB drive?</p>
            <p>I read in the faqs that <a href="https://www.apple.com/macosx/">OS X</a> programs may not run in opendarwin but console applications and x applications will. Fat use that is. This seems more primitive than the first version of linux (a <a href="http://www.slackware.com">slackware</a> distro) that I attempted to install so long (8 years) ago. I don't recall whether it had X on it (my installation certainly didn't) but it was certainly a lot easier to install and use.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

opendarwin

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>The PHP installation I described <a href="573.html">yesterday</a> had another twist, this time to do with the postgresql extension.</p>
            <p><a href="http://fedora.redhat.com">Fedora Core 3</a> distribution includes the RPMs for <a href="https://www.postgresql.org">postgresql</a>, the version is 7.4.6. Before <a href="000557.html">upgrading to fedora</a> I used to have a postgres 7.4.5 installed from source. Since I seemed to have doing nothing but <em>./configure, make, make install</em> for the past month attempting to recompile either version from source was ruled out, but the RPM installation does not seem to have the libpq-fe.h and as a result php will not compile:</p>
            <p><code>Cannot find libpq-fe.h. Please specify correct PostgreSQL installation path</code></p>
            <p>The trick is to install just the header files and not the full applications. To do this change into the <em>src/includes</em> and <em>src/interfaces</em> folders in your postgresql source tree. Then do a make and make install. Folow it up by ./configure --with-pgsql as usual in your php source tree and then a make and make install.</p>
            <p>That still leaves me with out the GD extension. When will this ever end?</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Cannot find libpq-fe.h.

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>When I compiled PHP after upgrading to Fedora Core 3, it was without two of my favourite extension, <a href="https://www.xdebug.org">xdebug</a> and <a href="573.html">turck cd mmcache</a>. I was in a hurry at the time but when I finally got around to compiling them both of them started throwing errors.</p>
            <p><code>configure.in:9: warning: underquoted definition of PHP_WITH_PHP_CONFIG</code></p>
            <p><code>run info '(automake)Extending aclocal'</code></p>
            <p><code>or see <a href="http://sources.redhat.com/automake/automake.html#Extending-aclocal">http://sources.redhat.com/automake/automake.html#Extending-aclocal</a></code></p>
            <p><code>configure.in:32: warning: underquoted definition of PHP_EXT_BUILDDIR</code></p>
            <p><code>configure.in:33: warning: underquoted definition of PHP_EXT_DIR</code></p>
            <p><code>configure.in:34: warning: underquoted definition of PHP_EXT_SRCDIR</code></p>
            <p><code>configure.in:35: warning: underquoted definition of PHP_ALWAYS_SHARED</code></p>
            <p><code>acinclude.m4:19: warning: underquoted definition of PHP_PROG_RE2C</code></p>
            <p><code>Can't locate object method "path" via package "Request" at /usr/share/autoconf/Autom4te/C4che.pm line 69, <gen1> line 111.</gen1></code></p>
            <p><code>aclocal: autom4te failed with exit status: 1</code></p>
            <p></p>
            <p>The link that's produced in the dump puts the blame squarely on a fedora modification to automake. In short Redhat wants everyone to create their make files the way they think it should be made.</p>
            <p>A simliar set of errors were produced for both xdebug and turck, but in the case of xdebug a more recent version (2.00 beta) was availabe, that produced the same set of warning but not the error.</p>
            <p>Finally I decided to give up on turck mmcache. It doesn't work well on php5 life is too short too bother learning m4</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

underquoted definition

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="http://subversion.tigris.org">Subversion</a> is full of surprises, whenever you start to think 'I am finally getting the hang of this' it throws you another riddle to solve.</p>
            <p>I ran into one of those riddles when I attempted to back and recover the repository. After having imported this web site, I followed it up with a few more projects that I am currently working on + some documents into the repository.</p>
            <p>That repository was dumped out again as instructed in the manual with the <em>svnadmin</em> command. (In case you want to know the exact command is: <em>svn admin dump /var/subversion --incremental &gt; svn-001 </em>). This produced a file of around 60MB - no great surprises there. The surprise came when the dump command was repeated today after making some minor modifications to the version controlled files.</p>
            <p>The new dump even with the --incremental flag was 80MB - now this is not what I expected, according to my reckoning it should have been closer to 20MB. This is the riddle in the subversion documentation.</p>
            <p>When you use the <em>--incremental</em> flag you have to recall memory (your's not the computer's) the highest revision number at the time the last backup was taken and start the dump from there. It would have been a lot better if the last revision number dumped was saved in the .subversion/ folder so that there is much less work to due. Apperently laziness is not considered a virtue everywhere.</p>
            <p>So far from the little I know of subversion it looks like the only way to effectively do is to call svnlook command to get information about revision numbers and then decide which ones to dump. So it may be a good idea to include the range of revisions as part of the dump file's name.</p>
            <p>As a parting shot it's worth noting that while the subversion docs contain a few riddles, the CVS manual is a riddle.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

subversion dump and restore

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Statements such as <a href="27.html">perl is better than PHP</a> spark off bloody religious wars among developers. In the admin world it's questions such as which distro is better or which file system is better that lead to divisions.</p>
            <p>As with programming languages, when it comes to file systems it is also a matter of chosing the right horse for the course or using the right club. The truth is for very small partitions ext2 is probably a lot better than either ext3 or reiser because it does not take up any space for the journal.</p>
            <p>When it comes to space utilization reiser is definitely better, the root partition of <a href="000557.html">my fedora core 3 installation</a> took up 6.0 GB when on an ext3 filesystem, it dropped to 5.5GB when it was moved to a reiser filesystem. That's a 10% difference.</p>
            <p>A couple of days ago I had a <a href="000565.html">corrupt partition table</a> on my hands but I still managed to recover everything on my ext3 partition with out any loss what so ever.</p>
            <p>Once upon a time I had a java desktop installation, java desktop as you know is based on Suse. It wasn't been used and needing space I removed it sometime ago and all the partitions were deleted. After the tables were corrupted, I made a new primary partion and instead of running <em>mkreiserfs</em> I accidentally typed <em>fsck.reiserfs</em>. Lo and behold a long forgotten installation of java desktop sprang up on me like a pheonix risen from the ashes.</p>
            <p>This reliability is nothing short of astonishing, that's perhaps why lots of mail server admins prefer reiser. Large installations of <a href="https://cr.yp.to">qmail</a> for example is IO intensive but it can run on a 100MHz CPU. In such cases, though reliability is essential, the extra cost of maintening the journal is too high. The solution then is to place the journal on a different device, which you can even do without formatting.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Religion Reiserfs vs ext3

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I updated to <a href=" http://www.movabletype.org/">movable type 3.121</a> a few days ago. It seems to have a bug, when I attempted to rebuild this blog, I started getting this error out of the blue:</p>
            <p><code></code></p>
            <p><code>Building entry 'Encrypted filesystem on linux' failed: Build error in template 'Date-Based Archive': Error in &lt;MTArchivePrevious&gt; tag: &lt;MTArchivePrevious&gt; can be used only with Daily, Weekly, or Monthly archives.</code></p>
            <p></p>
            <p>I certainly didn't make any changes to the templates or archive schemes that could have caused this error, after struggling with it for more than half and hour I finally realized that life is too short to bother with such trivial persuits. All I lose by removing the &lt;MTArchivePrevious&gt; and &lt;MTArchiveNext&gt; tag from the monthly template is that the pages no longer have links to the previous and next month but who cares?</p>
            <p>Interestingly, while upgrading MT, I decided (against my better judgement) to move from DBD to mysql as the backend. I would have moved to <a href="https://www.postgresql.org">postgresql</a> instead but the driver on <a href="558.html">my new Fedora Core Installation</a> kept choking, so regretably it was a mysql db that I turned to. (I can perhaps use, <a href="http://www.radinks.net/barbet/">Barbet</a> later on to migrate it again</p>
            <p>Since having the data stored in mysql, rebuilding the blog has become a lot slower than it used to be. Perhaps that is only to be expected, for semi static content there is nothing like a good old fashioned flat file and BDB is pretty close to a flatfile anyway.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Movable type bug?

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="../../megaupload">Megaupload</a> is once again back in the top 1% at <a href="https://sf.net/">Source Forge</a>. To be precise the stats for 4th December placed it at 99.42. As far as I am concerned that's a good thing, but how accurate are these stats?</p>
            <p>If you look closely the <a href="https://sourceforge.net/project/stats/index.html?report=last_30&amp;group_id=87646">stats page</a> is marked with a bug icon. That's the way it has been as far back as I can remember, perhaps an icon of a roulette wheel or a dice would have been more appropriate.</p>
            <p>Consider this: on the 4th Megaupload had 117 downloads and 965 page views. The stats are said to be calculated on a weekly basis, and the highest for the preceding seven days was 1901 pages and 160 downloads on the 1st of December. ON the strength of these figure the project was ranked 102 in terms of activity on the 4th.</p>
            <p>Now consider this, on the 20th of November, megaupload was ranked at lowly 1374, and 92 in terms of percentile. But take a look at the download and page view figures; 2029 pages were delivered on the 20th and 165 people downloaded the tarball. On the previous day 2102 pages were seen and 172 downloads took place. Despite these figures the project still slid down to 92%</p>
            <p></p>

        </div>
    </div>


</body>

</html>

The mystery of the Sourceforge Statistics

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I have been threatening to adapt <a href="http://subversion.tigris.org">subversion</a> for a <a href="343.html">long long time</a>. Finally I have taken the plunge by placing <a href="../../index.html">this site</a> into a subversion repository. What stuck me immediately was the fact that the size of the folder seems to have more than doubled after checking out. The increase with CVS is not so bad.</p>
            <p>It would seem that I have not correctly specified the path names because the checkout placed the files one folder deeper than I intended. Of course it's a simple matter to move it, but I tried to figure out how it can be moved with in the repository itself since that can become useful later on. You can move files with in the repository with the <em>svn move</em> command but it does not accept wildcards. So try someting like this:</p>
            <p><code></code></p>
            <p><code>for i in `ls`; do svn move $i .. -m'Imported to wrong folder'; done</code></p>
            <p></p>
            <p>You need to follow it up with a commit, it does lead to a slight swelling in the size of the repository but storage is cheap (even though <a href="562.html">my hard disks keep failing</a>)</p>
            <p>I was by now fooled into a false sense of security and tried to import a few other sites into svn repository as well. This time the files were placed one folder higher up than I intended. By now I had discovered a simple remedy to all ills that effect the repository: <em>rm /var/subversion -rf</em>, don't try this at office though.</p>
            <p>That rm command doesn't uninstall subversion but just cleans out the repository so that you can start a fresh, after a couple of fresh starts I did figure it out but not before realizing that new fsfs system - where the data is stored on flat files is superior (at least on this course) than the old BDB system. So another <em>rm, svnadmin create followed by svn import and svn checkout</em> sequence followed.</p>
            <p>In the meantime, I came to realize the the <em>svn propset</em> command syntax is not accurately described in the manual or I am too stupid to understand it so the only way for me to mark files to be ignored was to edit the ~/.subversion/config file.</p>
            <p>So to cut a long story short, I have finally 'got into' subversion as I have been threatening to do for a long time. A long time may pass again before i reach a conclusion as to which is better - subversion or CVS.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

From CVS to Subversion

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I had a sudden attack of paranoia and thought of placing some important stuff like source code and cryptographic keys on an encrypted file system. The <em>losetup</em> man page has a very good description of how this is done but there are couple of things to watch out for.</p>
            <p>The first one is that you need to load two kernel modules (if they have not been loaded already). The first of these is the <em>cryptoloop</em> module. Then you need to decide the encryption algorithm to use, it's said that <em>twofish</em> gives a nice balance between hard to crack and speed.</p>
            <p>On <a href="556.html">my Fedora Core 3 installation</a>, the exact steps to create a crypto filesystem of 256MB was as follows:</p>
            <p><code></code></p>
            <p><code>dd if=/dev/zero of=/mnt/bakup/swap bs=16k count=16k</code></p>
            <p><code>insmod /lib/modules/2.6.9-1.667/kernel/drivers/block/cryptoloop.ko</code></p>
            <p><code>insmod /lib/modules/2.6.9-1.667/kernel/crypto/twofish.ko</code></p>
            <p><code>losetup -e twofish /dev/loop1 /mnt/backup/crypto</code></p>
            <p><code>mkfs -t ext2 /dev/loop1</code></p>
            <p><code>mount /dev/loop1 /mnt/crypto</code></p>
            <p></p>
            <p>In your linux system you may not need to enter the full path to the module . You will need to have the loop devices in your /dev/ folder or you can create them with the <em>MAKEDEV</em> command (or equivalent) for your platform. In mycase /mnt/backup is the mount point for an existing partition, instead of dedicating an entire partition to the encrypted filesystem, I decided that it should be a file on an existing partition.</p>
            <p>Now the most important thing of course is not to forget the password that you choose when invoking losetup. After you unmount the file system you should call <em>losetup -d</em> on it. When you want to reuse it remember not to call dd and mkfs.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Setup an Encrypted filesystem on linux

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>If you attempted to boot a <a href="https://www.redhat.com">fedora core</a> installation out of a root partition of <a href="http://www.namesys.com">reiserfs</a> type you will most likely encounter an error, it may look something like this.</p>
            <p><code></code></p>
            <p><code>VFS: Can't find ext3 filesystem on /dev/hda1</code></p>
            <p><code>mount: error 22 mounting ext3</code></p>
            <p><code>...</code></p>
            <p><code>kernel panic - not syncing attempting to kill init.</code></p>
            <p></p>
            <p>With a little tweaking it's possible to boot your Fedora Installation off a reiser filesystem, The most important thing to realise is that reiserfs is compiled as a module and not part of the main <a href="https://www.kernel.org">kernel</a>. That would have been ok, if the default ram disk does include this module, apparently it does not.</p>
            <p>1) edit the fstab</p>
            <p>Once you boot up your system using the rescue mode (or using a live cd like damn small linux or knoppix), you need to edit your fstab to ensure that the '/' entry points to the correct location and that the fstype is set to reiserfs.</p>
            <p>2) Create ram disk.</p>
            <p>Look in /lib/modules/ to find out what version of linux you are using. In the case of Fedora Core 3 it is 2.6.9-1.667. Then you can create a ram disk by invoking the mkinitrd command. (<em>mkinitrd initrd-2.6.9-1.667.img 2.6.9-1.667</em>)</p>
            <p>3) Boot manager.</p>
            <p>Copy the image produced above into the boot folder under a different name and modify your existing record in grub.conf to use the new ramdisk. For example in my case the initrd entry was changed to: <em>initrd /boot/initrd-2.6.9-1.667.2.img</em></p>
            <p>4) grub-install</p>
            <p>The easiest way to updat the boot manager is to use the <em>grub-install</em> command. In my case i use <em>grub-install /dev/hda</em></p>
            <p>5) reboot and enjoy</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Boot Fedora from a reiserfs root

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>This is the story of a corrupt partition table recovery in linux, a story with a happy ending.</p>
            <p>The hard disk in question was a 40Gg Maxtor, after one of it's partitions was deleted <em>fdisk</em> started reporting a partition which was well in excess of the disk size. The result was a system that refused to boot, recovery proved to be quite a strugle.</p>
            <p><a href="556.html">Fedora Core installer</a> only gives you the option of choosing between ext2 and ext3 for your filesystem. The better journaling file system for linux however is reiserfs. So there I was pruning and grafting my partitions trying migrate them from ext3 to reiser when the partition table became damaged.</p>
            <p>As a first step <em>parted</em> was tried and failed because it refused to work with a slice that was larger than the physical disk size. The second attempt involved <a href="http://www.stud.uni-hannover.de/user/76201/gpart/">gpart</a>, which did look the part. After a scan it produced a report that resembled the last valid partition table but I didn't have the confidence to proceed with it because gpart does not officially support ext3.</p>
            <p>However gpart did give me enough information to realise that if I deleted the offending partition, which appeared after all others, I might be able to recover the rest. So it was fdisk that was called up again to delete the unwanted partition, that didn't solve all the problems because the so called secondary partition now appeared to be flattened and it's individual slices (logical drives) were not showing up.</p>
            <p>parted was called on once again, this time to attempt to recover the logical drives. The approach is to run the rescue command of parted. It requires a start and end parameter (both in megabytes). It's safe to use the end value for the primary partition for the start parameter, the end parameter has to be guessed from the approximate size of the hard disk.</p>
            <p>This is the approach I adapted after increasing the ending value by small amounts and running the rescue command several times it finally detected the frist slice with in the secondary partition which I promptly recovered and <em>fsck</em>ed. The rest I don't really care about and abandoned it.</p>
            <p>Of course both these programs were exectued after booting the system with a rescue cd. Instead of a rescue CD you could use a live linux cd such as <a href="http://www.knoppix.org/">knoppix</a> if one is avaialable. Not writing anything at all to the hard disk is crucial if you want to recover your data, since knoppix and rescue CDs write their data to a ram disk they are safe to use.</p>
            <p>The reiserfs home page is at <a href="http://www.namesys.com">http://www.namesys.com/</a>, it's amazing how many sites errornously link to http://www.reiserfs.org instead. Be warned that the above site takes an eternity to load if you are on a slow connection.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Corrupt Partition Table Recovery.

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Finally I have a few nice things to say about <a href="https://www.redhat.com">Fedora</a>. The most impressive thing about it is the fact that almost all it's packages are so spanking new. For example it uses <a href="http://www.gnome.org">Gnome 2.8</a> and it seems like yesterday when I struggled with a <a href="503.html">manual installation of gnome 2.6</a>.</p>
            <p>As related change, it now uses Xorg instead of XFree86, one of the problems that effected my compilation of gnome from source apparently had to do with a couple of issues in XFree86. All these changes to X, gnome and KDE also means that drag and drop is more consistent and yes you can now <a href="http://www.radinks.com/upload/dnd.php">drag and drop upload files</a>.</p>
            <p>I also like the fact the CD Burning is no longer the painfull affair it used to be, but this is thanks mostly to <a href="http://www.k3b.org">k3b</a> more than anything else. The same can be said about printing, in my old distro some of the fonts looked absolutely hideous when printed. While on this topic does <a href="https://www.openoffice.org">open office</a> really start up a lot quicker or do I feel that way because of <a href="562.html">my new machine</a>?</p>
            <p>Now for the brick bats; Why is it that <a href="https://www.mozilla.org">the Mozilla browser</a> now refused to start up untill I installed the backward compatibility libraries? they should have been installed by default. Most software is still designed to be linked against older versions of libc.</p>
            <p>The biggest gripe though is with SELinux. SELinux apparently thinks it's ok to copy files from one folder to another and to delete the source - which is of course the equivalent of moving files, however it does not allow you to use the mv command in this way. This statement should be qualified by saying the super user was moving files that belonged to one user to another folder that was owned by a different use, but still if mv is disallowed the same should apply to cp as well.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

The Fedora Core Experience.

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>It's been two weeks since my last post, I have been down with indigestion caused by Mother Board Stew.</p>
            <p>Immidiately After I completed my migration from Red Hat 9 to Fedora Core, one of the three hard disks on my computer fell criticially ill and died. All that whirring about while Fedora was being installed proved too much for. The death of the hard disk was no great loss but the time lost in installing Fedora Core several times proved to be too much and I had to spend a lot of my time with other work just to catch up with the back log.</p>
            <p>It turned out that the hard disk failure was caused by a mother board problem. In <a href="https://www.raditha.com/archives/2004_03.html">March</a> I wrote about a suspected <a href="52.html">toast CPU</a> and I have encountered other niggling problems like <a href="000522.html">spontanous restarts</a> the real culprit as it now turns out is the Mother Board.</p>
            <p>Nearly an year ago, me and a friend both purchased Gigabyte ----- motherboards and set up nearly identical systems. Geeks in Sri Lanka are in the habit of setting up their own machines, this may be true everywhere but in Sri Lanka we have an added incentive because branded machines are twice as costly. You find many computer shops selling cheap unbranded machines but walking into one of these shops feels like walking into a used car dealership.</p>
            <p>My friends mother board lasted less than 3 months the vendor did replace it but it died once again and took his CPU along with it to silicon heaven. In my case after the hard disk burn out, I plugged in a new one but the MB absolutely refuse to detect it. To be precise the IDE controller refused to work with four devices (3 Hard Disks, 1 CD Writer) at the same time. It was then that I finally concluded that the mother board was stewed.</p>
            <p>I have been using AMD processors for a long long time, over the last few years I have been through a K6-II, K6-III, Athlon and now an Athlon XP. All four had to be discarded before their time was up because the mother boards were toasted. Thus for a change I decided to try an Intel primarily because Intel makes it's own boards. So for the first time since working on a Pentium 166 many many years ago, I now have another Intel chip ticking away inside my computer.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Mother Board Stew

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Yesterday when I finally managed to boot up after <a href="000557.html">migrating from Redhat 9.0 to Fedora Core 3</a>, it was with a minimalistic installation. An installation without <a href="http://www.gnome.org">gnome</a>, <a href="https://www.kde.org">KDE</a> or any of the office or developer tools. Adding these packages turned out to be a wrist spraining affair.</p>
            <p>The package manager at first refused to recognize CD changes, this was tracked down to the fact that the cd was mounted at /media/cdrecorder and not /mnt/cdrom creating a new folder in /mnt seemed to take care of that but I had to change CDs at least 30-40 times to get all the packages in.</p>
            <p><a href="https://mozilla.org">Firefox and thunderbird</a> when they were installed did not detect the old settings properly so I decided to browse with konqueror and download the latest versions of these software. The download manager for kde, Kget seemed to work fine for a while but then it just disappeared and nothing would coerce it to show up on the desktop again so I had to go back to the versions found on the CD. (yikes it does not recognize the <a href="545.html">java console extension</a>)</p>
            <p>Thunderbird no longer appeared to choke with the <a href="000549.html">Segmentation fault</a> I reported earlier but it had the same problem all previous versions of mozilla mail had; <a href="458.html">it chokes when copying a message to the outbox</a></p>
            <p>Having lost a full day thanks to the installation/upgrade process, I was desperately keen to get back to work and decided to take a risk and install <a href="https://www.apache.org">apache</a> and PHP from RPMs, I should have known better. It was a big mistake. The news installation refused to execute cgi scripts outside the cgi-bin folder. The log file says <code style="color: rgb(0, 0, 102);">Permission denied: /home/raditha/websites/raditha/mt/.htaccess pcfg_openfile: unable to check htaccess file, ensure it is readable</code></p>
            <p>There was a lot of bull in various mailing lists about this having to do with file system permissions. That's most unlikely, after working with apache for around 8 years i am sure i would know how to set the correct permissions. Rather than wasting time trying to figure this out (you are not supposed to use RPM installations of apache anyway). So apache was compiled from source along with a minimalistic php engine. Hopefully I will be able to get back to work quickly and have something good to say about <a href="https://www.redhat.com">fedora</a> tomorrow.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Red Hat 9 -&gt; Fedora Core Part II

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="556.html">Yesterday</a> having successfully managed to install <a href="https://www.redhat.com">Fedora Core 3</a>, I decided to take the plunge and upgrade my heavily patched Red Hat 9.0 installation. It ended up being a disaster.</p>
            <p>Things got off to a bad start when the installer kept insisting that my /usr partition was too small. That is a lie, there was more than 1.1 GB free on it, but the installer wanted 430MB more. I cleaned out some unused files and moves some others to a different partition and managed to free up around 500 Mb. Then the installer wanted 19MB more!!.</p>
            <p>What's really annoying is that you get this message after traversing quite a distance down the installation route. You then have to reboot your existing installation, free up space and then boot up again with the installation CD.</p>
            <p>After quite a few attempts the installer managed to get over this disk space stumbling block but to my dismay I found that it will take around 150 minutes for the upgrade. The fresh installation took only about 40 minutes. After about 125 minutes the installer freaked out claiming one of the packages were corrupted. The end result was that my existing installation was messed up beyond repair and I didn't have anew installation to show for my efforts either.</p>
            <p>Booted up again with the installer and this time took to the hard disk partitions with a butcher's knife. I cleaned out the /usr and/usr/local partitions and made them into a stripe set. Then started afresh installation, which did complete in about 40 minutes. That's not the end of the story. When I finally managed to boot up again, it was only to find that the old root and the /home partitions were totallycorrupted this definitely is time to take a break.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Red Hat 9 -&gt; Fedora Core

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            Just got hold of a set of Fedora Core 3 CDs and had a go at installing it on my machine that is currently running Red Hat 9.

            This is the first time that a Fedora CD is seeing the inside of my CDROM. That's because I have a highly customized version of Red Hat 9.00. Customized in the sense that very few of the packaged originally installed by way of RPMs now survive. Most have been replaced by packages compiled from source.

            <p>Getting back to the installation; I chose to make a fresh installation instead of an upgrade by pluggin in a hard disk that was lying around. The fact that the Reiser file system is not given as an option when partitioning is a big disappointment. That means I would have to manually convert to reiserfs later on. ext2 and ext3 cannot
                hold a candle to reiser.</p>

            <p>I have a 16MB partion on my primary hard disk (which has seen three different CPUS) that I use as the boot partition. Long long ago you had to have a primary partition on your primary hard disk for many linux distros to be able to boot. It was this partition that I set as the hard disk (there is still 8MB free on it) and the installer suggested that I increase the partition to 75MB. What for the kernel is only about 2Mb, so this warning was ignored.</p>

            Going through the packages I noticed with great delight that <a href="481.html">festival</a> is part of fedora core 3. oh boy. (In case you have not run into it before festival is a TTS (text to speech) engine)

            Finally I noticed a new feature SE Linux which apparently has to do with security in the same page as the firewall configuration screen. I have to hang my head in shame and proclaim that I have not heard of SE Linux till now.

            Having sorted things out and choosing all the packages I started the installer and grabbed the newspaper. It went through three cds (a process that took more than 40minutes) to finally announce that the Korean language pack could not be installed. The guy who wrote the installer forget to put a cancel button on that dialog. It only has an OK button, and when you click on it, the installer will attempt to install the package and keep doing this until you hit control alt delete.

            So I had to do the whole thing all over again this time with out the language packs and fedora started up alright but I had to edit the grub configuration file to get back to my frayed at the edges Red Hat 9 with patches on patches.


        </div>
    </div>


</body>

</html>

Fedora Core 3 Installation

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="https://www.java.net">java.net</a>, the so called <em>source for java technology collaborartion</em> has a sense of humor. We already have <a href="http://java.sun.com">java.sun.com</a> and <a href="https://www.java.com">java.com</a> confusing enough for most people even without all the subsections on java.sun.com</p>
            <p>So what exactly is java.net? it looks a like a collection of blogs, and news feeds which is another way of saying this is just another website. Hang on a second they also host a few projects and have CVS servers etc for developers. Sure but would you move from <a href="https://sourceforge.net">sourceforge</a> just because of that?</p>
            <p>Now to look up the real humorous items on their website; the FAQ says you could 'federate your site' with java.net. What does federating mean? <b></b></p>
            <p><b>
                    <quote>""To federate" means, "To cause to join into a league, federal union, or similar association." (American Heritage Dictionary). On java.net, federated sites will typically be sites that are already mature. They have an existing, stable community and are self supporting. For a variety of reasons, these groups may wish to maintain</quote>
                </b></p>
            <p><b>their own established infrastructure and to tie into some of the enriched features that java.net offers"</b></p>
            <p><b>If you don't want to federate (probably because you cannot see the benefits of doing so). You can ask java.net to link to your website (from a page deep inside their cavernous site) in return you have to link to java.net and place their logo on your own home page. What a nice bargain. I remember websites making these demands around eight</b></p>
            <p><b>years ago when I first started browsing. Maybe this site was made eight years ago and left to rot.</b></p>
            <p></p>

        </div>
    </div>


</body>

</html>

java.net is pulling your leg

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>We have been threatening to release the <a href="http://www.radinks.com/print/">Rad Inks Print Suite</a> for a long long time. Then on august 31, we finally did. On November 04th we followed it up by the first new version (1.01).</p>
            <p>There aren't that many significant feature enhancements in the new version (except for the fact that a single order can now have multiple files associated with it). The most significant change is that you can now download the software free of charge for <a href="http://www.radinks.com/print/eval.php">evaluation</a> purposes.</p>
            <p>Since we have placed a price tag of $1950 on the release version we are taking quite a risk but we firmly believe that our clients are honest and decided to go ahead with the free download.</p>
            <p>We have been running our online store for over an year. During that time we have had only one (yes 1) charge back. The percentage of orders we have had to cancell due to fraud has been very small. That's why we have a lot of faith in the customer's honesty.</p>
            <p>And what about the <a href="http://www.radinks.com/print/card-designer/">card designer</a>? Version 1.10 of the card designer was released today. The main purpose of this release is to ensure that the card designer and the card</p>
            <p>designer embedded into the print suite are in sync.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Print Suite and Card Designer

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Though I have been a programmer for7 years I have never quite got the hang of date/time calculation in any of the programming language I know (I can assure you there are quite a few). Recently I needed to calculate the time elapsed since a thread was started and here is what I came up with first:</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Date and Time Difference in Java

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Recently a client who ordered a customized version of <a href="http://www.radinks.com/upload/">Rad Upload</a></p>
            <p>had a problem with the applet when it was deployed on his server. It worked fine when it was run off our own servers. The error was something like this:</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

javax.imageio.spi.ImageWriterSpi: Illegal configuration-file syntax

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I decided to find out how many duplicate accounts exists in one of our databases and made a startling discovery about mysql sub queries.</p>
            <p>For many years PHP scripters (and other programmers for that matter) have been using a nested loop in place of subqqueries in mysql. The enlightened decided not to bother with mysql at all and to rely on postgres instead. Not very long ago mysql finally decided to implement subqueries but it is still not all tha t prevalent because many ISP</p>
            <p>still support older versions of mysql which does not support subqueries, foreign keys and many other features.</p>
            <p>To get back to my problem, I reconed that if anyone had signed up using the same email twice that would be a duplicate account. This was on a Rad User Manager database that had around 1600 users at the time. I used the following query to identify duplicates.</p>
            <p>select a.userId, b.userName, b.userPassword, a.userEmail, a.userValidated</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

mysql subqueries

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="http://www.ca.com">Computer Associates</a> seems to think that software is used only in a handfull of countries. At least that is the impression you might get if you read up on their so called '<a href="http://www3.ca.com/Solutions/Collateral.asp?CID=61137">Million Dollar Challenge</a>.</p>
            <p>The pdf that describes this contest says in bold capital letters:</p>
            <p><code>THE CONTEST IS INTENDED FOR PRESENTATION IN THE UNITED STATES, CANADA (EXCEPT QUEBEC PROVINCE), MEXICO, INDIA, CHINA, THE UNITED KINGDOM, AUSTRALIA AND NEW ZEALAND. DO NOT PROCEED WITHIN THIS SITE IF YOU ARE NOT A RESIDENT OF ONE OF THESE COUNTRIES.</code></p>
            <p>This is a company that was once headed by a CEO from <a href="http://www.saadhu.com/srilanka/">Sri Lanka</a>. Sri Lanka isn't in that list of countries either. What about european countries? Singapore, Malaysia? they are all conspicous by their absence.</p>
            <p>I am not surprised that CA needs to spend 1 million dollars to persuade open source developers make tools that will migrate from other databases to ingres. This $1,000,000 probably does not include the advertising budget.</p>
            <p>You would think a software company would know how to filter their banner ads by region, but they do not because I came to know of this 'challenge' through an ad on <a href="http://www.freshmeat.net">freshmeat</a> (or was it <a href="https://www.sf.net/">SF</a>?). Where do I live? Sri Lanka Of course.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

million dollars wasted?

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Having installed firefox a few days ago, I decided to have a go at installing thunderbird. The firefox installation certainly didn't go very smoothly for me, which was disapointing.</p>
            <p>Unlike the other products I try out on an almost weekly basis, firefox been a browser has every single internet user as it's target market - not just software developers like you and me. As such it should be really smooth sailing, which was not to be.</p>
            <p>Thunderbird installation was a lot smoother, it even managed to import all the mail from my mozilla 1.7x installation. I have been fed up with mozilla mail for qutie some time. The most annoying thing about mozilla mail is that it frequently fails to copy messages to the sent box, even if the sent box is a local folder.</p>
            <p>Occaisionaly I have had reason to believe that mozilla may be failing to deliver mails altogether. That's because couple of people I work with have claimed not to have recieved emails I sent them. And lo and behold there isn't a copy in the outbox.</p>
            <p>So all in all there were plenty of good reasons to move to thunderbird. It looked quite promising until I attempted to reply to a mail. Thunderbird just disappeared from my screen with the message:</p>
            <p><code></code></p>
            <p><code>/usr/local/thunderbird/run-mozilla.sh: line 451: 3936 Segmentation fault "$prog" ${1+"<a href="../../cdn-cgi/l/email-protection" class="__cf_email__" data-cfemail="250165">[email&nbsp;protected]</a>"}</code></p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

thunderbird

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I have written a few times before about bugs in Java 5 (see <a href="534.html">Oct 12th entry</a> for example). A more serious bug something which I have often noticed before but not blogged about is the class verification error that is often seen by many a developer (and user of java 5.0). Here is a detailed example of the error (minus the full stack trace)</p>
            <p><code style="color:#000066">Exception in thread "main" java.lang.VerifyError: (class: com/radinks/VerifyTest, method: rename signature: (ILjava/lang/String;)Z) Incompatible object argument for function call</code></p>
            <p>A class that runs perfectly well in previous versions of java will obstinately refuse to load in java 5.0. I had dismissed this previously as an issue in the beta, but the beta is long gone so it really should have been resolved.</p>
            <p>What's astounding is the fact that this problem has been reported as early as april and it's still not fixed. (ref <a hreaf="http://forum.java.sun.com/thread.jsp?forum=481&amp;thread=514766&amp;tstart=0&amp;trange=15">http://forum.java.sun.com/thread.jsp?forum=481&amp;thread=514766&amp;tstart=0&amp;trange=15</a>). This is a real show stopping bug. Many application or applet that was compatible with older versions of the JRE will simply refuse to function with java 5.</p>
            <p>The thread above has a suggestion that will work as a temporary work around pass the '<em>-noverify</em>' parameter to the jvm. As developers you and I can do that but don't expect the end user of your application to do the same.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Incompatible object argument for function call

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>It was <a href="531.html">only three weeks ago</a> that I made the previous release of the PHP edition of <a href="../../megaupload">mega upload</a>, It had to be followed up by another release today because 1.44 is not as secure as it should be.</p>
            <p>As a downloader named Dimitry pointed out, the 1.4x branch passes the list of file names as part of the querystring. It does not make use of the internal post as the 1.3x branch does. This unfortunately leads to the possibility of a malicious user passing in bogus pathnames that could result in files been overwritten. Or would allow the user making a copy of important files on the server.</p>
            <p>It should also be noted that the 1.4x branch readme file clearly states: <b>MEGA UPLOAD 1.4x IS EXPERIMENTAL PLEASE DON'T USE IT ON PRODUCTION SITES USE 1.35 INSTEAD. 1.35 IS STABLE.</b></p>
            <p>Well the issue has been addressed and in doing so another improvement has been made which makes it possible to handle a larger collection of files than the previous version. Here the reference is not to the file sizes but the number of files handled at a single upload.</p>
            <p>Finally always remember that megaupload is just a progress bar for file uploads nothing else. You need to treat files uploaded with megaupload with the same level of distrust that you would with ordinary file uploads.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Megaupload

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Many moons ago, <a href="343.html">I downloaded subversion</a> because I had messed up the CVS installation on my computer (not that I was making much use of it anyway) and wanted to try something different. Finally I had a chance to install it.</p>
            <p>One of my worst habits is downloading a software and leaving it to rot in some remote corner of my hard disk. While searching for something or other I stumble upon these and decide to have a go at instalation. Often the version number would have changed since the time I downloaded but it doesn't really matter, I am usually installing software just for the sake of finding out what it is.</p>
            <p>Subversion apparently depends on the <a href="https://apr.apache.org/">apache portable runtime </a>&nbsp;(apr), judging by the time taken by the configure script to complete, that's just the tip of the iceberg. I started installing mono a couple of days back and that certainly seemed to go much faster.</p>
            <p>I immidiately moved the entire folder to /dev/shm before continuing with the installation. Othewise this might take as long as the kernel to compile itself. After that the compilation tooks just a couple of minutes.</p>
            <p>Once the installation is complete the next step was to create the repository, which was done with <em>svnadmin create /var/subversion/</em> before importing any projects I had to spend some time on RTFM, since my project folders contain compiled classes and other stuff which should not make be placed into the repository. In fact the area of TFM that covers svn:ignore is very poor. In fact the subversion manual is almost as bad as the CVS manual.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Subversion

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>After spending quite a bit of time <a href="544.html">installing firefox</a>, the next step is to try out an applet on it. So I made the link to <code style="color:#000066">jdk1.5.0/jre/plugin/i386/ns7/libjavaplugin_oji.so</code> and started up the browser again, it worked despite the fact that java 5.0 has quite a few problems of it's own.</p>
            <p>Hang on where did the java console go? It turns out you need to install an <a href="https://update.mozilla.org/extensions/">extension</a> if you want to access the java console from <a href="https://mozilla.org">firefox</a>, fortunately the extension is only 7 kb so it installed in seconds. The next victim (obviously) is <a href="http://www.radinks.com/">rad upload</a> and guess what it worked perfectly.</p>
            <p>Now the symlink in the plugins directory was changed to the 1.4.2 version (which is probably the most stable version of java ever) and the same process was tried with success.</p>
            <p>Then I had this sudded urge to try out the <a href="http://www-106.ibm.com/developerworks/java/jdk/linux140/">IBM JVM (v 1.4)</a>. This was one of the more popular JVMs for linux systems for a long long time but has waned in popularity in recent years. I already have a 1.3 version installed but wanted try out the new version as well. Firefox passed that test with flying colors too, the most tedius part was to fill up all the forms on the IBM developerworks site.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Firefox and java

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I wrote a couple of days back about <a href="539.html">trouble installing firefox</a>, the error message <code style="color:#000066">./firefox-installer-bin: relocation error: /usr/local/lib/libgtk-x11-2.0.so.0: undefined symbol: pango_find_base_dir</code> wasn't in google's index.</p>
            <p>There were less than 10 results for the second half of that error message <code style="color:#000066">relocation error: /usr/local/lib/libgtk-x11-2.0.so.0: undefined symbol: pango_find_base_dir</code>. Of course these were related to some other applications and not to firefox, but one of them seemed to indicate that having two installations of pango could be the problem.</p>
            <p>I did have the RPM of pango-1.2.1-3 as wel as pango-1.4.0 installed from source (as part of my (botched) attempt to upgrade to gnome 2.6). At the time I didn't uninstall pango-1.2.1-3 because it had a list of dependencies as long as your arm. Now is the time to take the plunge, or so it seemed.</p>
            <p>After uninstalling the RPM, saving this entry and shutting down mozilla, i tried the installer again. This time it went a few steps before grinding to a halt with a different error.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Firefox and xpistub

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I am beginning to wonder why I ever <a href="517.html">bothered with RAM disks</a>. Apprently tempfs can offer all the benefit of RAM disk and more.</p>
            <p>For years I used to think /dev/shm that used to show up with a call to <em>df</em> was just for shared memory. How stupid can you get?! according to the kernel documentation, it probably isn't even used for shared memory nowadays!</p>
            <p>It turns out that you can mount /dev/shm and use it just like any other partition. The main advantage it has over RAM disks is that you can change it's size dynamically and if by some chance you need it to be larger than the available RAM you can move it out into swap.</p>
            <p>The first thing I did since making this discovery is to move my /tmp into shared memory. All it takes is the following entry in /etc/fstab</p>
            <p><code style="color:#000066">none /tmp/ tmpfs size=32M,nr_inodes=10k,user,noatime 0 0</code></p>
            <p>That gives me a 32mb /tmp folder.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

tempfs vs RAM Disk

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Just downloaded the latest version of MovableType (3.11). They certainly have made it very difficult to download it. They have one of those wierd images with distorted set of characters which you have to read out and enter into the form to complete your download. It took me three attempts.</p>
            <p>I probably shouldn't be complaining because I am using the free version anyway. Movable type is a very good software simple but versatile, just the way I like it. If I ever find myself with a bit of cash to spare I am going to buy the paid version, not because I need it (this is a personal site) but because I think they deserve the support.</p>
            <p>Now to get on to the upgrade; I am moving from 2.661 to 3.11 they have a special download for upgraders and that is what I obtained and untarred. The blog itself has been tarred up and moved to a safe place ready for the upgrade, when we continue we will be using the new version.</p>
            <p>The upgrade script mt-upgrade30.cgi choked:<code style="color:#000066"></code></p>
            <p><code style="color:#000066">Upgrading your databases:</code></p>
            <p><code style="color:#000066">An error occurred while upgrading the schema:</code></p>
            <p><code style="color:#000066">Tie to 'db/author.created_on.idx' failed: Permission denied at /home/raditha/websites/raditha/mt/lib/MT/ObjectDriver/DBM.pm line 610.</code></p>
            <p></p>
            <p>This was overcome by changing the permissions on the db folder to 777 something which was not needed with the previous version. I decided it might be better to change ownership to nobody and change the permissions back to 700 before running the mt-upgrade31.cgi script.</p>
            <p>I am still faced with a minor hiccup, I have forgotten my password and this installation is on my desktop, a machine that does not have a mailserver (well I did have qmail but a recent hard disk error killed it. Does that mean my password is lost for ever? Not really, I just edited the recover_password method in CMS.pm to echo out the password instead of mailing it.</p>
            <p>That's when I realised that I had messed up my permissions altogether because MT was claiming that the username was not found. Changing the perms again sorted things out.</p>
            <p>The last few paragraphs above were added using the new version. The UI for the old version was heaps better than this.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Movable Type Upgrade

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>After installing the <a href="../../wiki/index.html">wiki</a> a few days ago I decided to change the appearence of the entire site to match the wiki. Of course there is considerable risk involved, UI changes more often than not result in sites being confined to obsecurity because they lose what ever positions they may have held in SEs.</p>
            <p>While on the topic, a couple of days ago, when I logged into windows after some time, I noticed that another Page Rank calculation had been carried out. Not sure when exactly the recalculation took place but according to my reckoning it's been <a href="414.html">at least three months</a>. I guess the long anticipated three monthly cycle is finally upon us. Don't be suprised if many blogs and mailing lists archives are going to loose the ranks they enjoy now.</p>
            <p>As for the UI change, this is what the site looked like until Oct 15th.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Site Overhaul

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>The newly installed wiki caused the cgi scripts on this site to stop functioning. The actual culprit was the URL rewrites that were used with wiki to translate 'ugly urls' (as mediawiki authors call it) into the 'good looking' URLs that you see on the <a href="../../wiki/index.html">Wiki</a> section of this website. The exact error was :</p>
            <p><code style="color:#000066">Options FollowSymLinks or SymLinksIfOwnerMatch is off which implies that RewriteRule directive is forbidden</code></p>
            <p>This error unforunatley made it impossible to access the online demo for the <a href="../../megaupload">megaupload</a> project for a few days (I added the wiki a day after the new release and didn't bother to test the perl scripts afterwards)</p>
            <p>Solutions were penny a dozen a when you searched for that error in google, unfortunately many of those solutions had apparently not worked for the original posters. The common consensus was that if your CGI stopped working you need to add the <em>Options +FollowSymlinks</em> to your httpd.conf or .htaccess .</p>
            <p>In my case adding the Option to my .htaccess file simply did not work (as many others had found before me, (many of them appear to have given up all together), it seems the directive has to go into the httpd.conf file - or to be more precise it needs to be placed with in the &lt;Directory/&gt; tags as follows:<code style="color:#000066&quot;"></code></p>
            <p><code style="color:#000066&quot;">&lt;Directory "/home/raditha/websites/raditha/cgi-local/"&gt;</code></p>
            <p><code style="color:#000066&quot;">AllowOverride None</code></p>
            <p><code style="color:#000066&quot;">Options +FollowSymlinks</code></p>
            <p><code style="color:#000066&quot;">Order allow,deny</code></p>
            <p><code style="color:#000066&quot;">Allow from all</code></p>
            <p><code style="color:#000066&quot;">&lt;/Directory&gt;</code></p>
            <p></p>
            <p>Just in case you think me stupid for sharing part of my configuration file, this is not from the server but from my desktop.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Rewrites and CGI

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>After a lot of effort I finally manage to mount <a href="451.html">my Creative MuVo Tx as a USB drive</a>. As it turns out, part of the problem was that I had turned off SCSI support. According to what I gathered you don't need SCSI emulation for CD burning if you are using linux 2.6 (I haven't upgraded my CD Burning software yet so i haven't had a chance to try this out yet (CD burning certainly is an area where linux distros have been lagging behind)</p>
            <p>While browsing I came to know that some USB devices may not be detected if they are cold plugged, the same can happen if you configuraiton is not correct. So I unplugged and plugged in my thumb a couple of times while keeping an eye on <em>/var/log/messages</em> it threw out some errors produced by <a href="http://linux-hotplug.sourceforge.net/">hotplug</a>. So I decided to upgrade hotplug since the version on my machine is more than 2 years old.</p>
            <p>Then quite fortunately I came across this link: <a href="http://sxs.thexdershome.com/storage/clik_usb.html">http://sxs.thexdershome.com/storage/clik_usb.html</a> and that's how I learnt that not compiling the SCSI driver may be the root of all evil.</p>
            <p>Now my USB drive shows up as /dev/sda1 and it mounts itself as a vfat drive but I am happy. Hopefully my webcam will spring to life pretty soon. Now to figure out why <a href="523.html">my 2.6.8 kernel</a> runs a lot slower than <a href="319.html">the 2.6.5 version</a>.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

USB Drive on linux 2.6.8

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="https://www.mozilla.org">Mozilla</a> is getting further and further up my nose everyday so instead of upgrading to mozilla 1.73 I decided to try out firefox instead. This is not the first time that I am trying out firefox the first attempt was about an year ago but at the time it had quite a few problems with applets so it fell victim to an `rm`.</p>
            <p>Mozilla is too smart. Try typing in <a href="http://www2.radinks.com/">www2.radinks.com</a> into your browser chances are it will out think you and take you either to <a href="http://www.radinks.com/">www.radinks.com</a> or to <a href="http://radinks.com/">radinks.com</a> the problem here is that the three urls point to three different servers.</p>
            <p>There is often a need to try out all three urls given above, to make sure that they are in sync, unfortunately mozilla insists on taking you where it thinks you should be going not where you want to go. If that's the way I want my software to behave I will work on windows.</p>
            <p>The other major irritation with mozilla is that it frequently fails when copying a mail that has been delivered to the sent items folder (outbox). Since upgrading to gnome 2.6 the error message <code></code></p>
            <p><code>Gtk-WARNING **: libgdk_pixbuf.so.2: cannot open shared object file: No such file or directory</code> always pops up when you try to start up the browser.</p>
            <p>Enough about mozilla now for firefox. This is what happens when I try to install firefox:</p>
            <p><code>./firefox-installer-bin: relocation error: /usr/local/lib/libgtk-x11-2.0.so.0: undefined symbol: pango_find_base_dir</code></p>
            <p>Certainly not the most auspicious of starts but I am willing to give it the benefit of the doubt and put the blame on my gnome installation.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Mozilla and Firefox

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Hard On the heals of the <a href="http://www.radinks.com/upload/">Rad Upload</a> release comes a new version of the <a href="http://www.radinks.net/user/">Rad User Manager</a>. This is one of the handfull of open source projects that Rad Inks is involved in. (<a href="../../megaupload">Mega upload</a> is not one of them, megaupload is a personal project)</p>
            <p>The new version gives you the choice of storing your PHP session data in a database, the old approach of saving this information on the session files at the server is still supported and continues to be the default setting.</p>
            <p>Rad User manager does not store anything critical information in cookies. The only thing that will be saved in a cooky is the session identifier, the data associated with the session stays safely on the server with out being passed back and forth between the browser and the server.</p>
            <p>If you don't want to save even the session identifier in a cookie you need to disable cookies in your PHP configuration (php.ini) file and enable trans-sid. In some cases you might need to compile PHP with --enable-trans-sid support. Some developers are of the view that trans-sids are less safe than storing the session id in a cookie. You should do your own research before deciding what option is best for you.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

User Manager

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>There was a new release of <a href="http://www.radinks.com/upload/dnd.php">Rad Upload</a> yesterday, this is th first new release after a period of more than 2 months. The new release <a href="http://www.radinks.com/upload/changes.php">version 1.30</a> adds support for SSL encryption. That means Rad Upload can now be used with HTTP, HTTPS or FTP.</p>
            <p>The release almost coincides with the 1st anniversary of the product. It was first release in October of last year and we had the first sales on Oct 14, 2004 we have come a long way since then. I like to think that we have played a significant part in the gradual comeback that is being made by applets.</p>
            <p>When we started out with <a href="http://www.radinks.com/sftp">our SFTP applet</a> applets being used for file management was extremely rare and ours was the first SFTP applet. Now there is stiff competion which we welcome.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Rad Upload and HTTPS

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Did you see that <a href="000532.html">really massive stack trace</a> (that was thrown by the supposedly stable version of java 5) in my blog a few days ago? I think I may have found the solution.</p>
            <p>If you recall the java plug in seemed to have a problem at least in java 5 where a problem did not exist in the beta version. The choker is a little bit of style. The server side upload handler that we include as a sample with <a href="http://www.radinks.com/upload/">Rad Upload</a> includes a little of bit of style hich is valid.However we do set a style of 110% to one of the table headers, java 5 seems unable to process it, even though previous versions could.</p>
            <p><code> the culprit: &lt;td style="font-size: 110%" align="right"&gt; </code></p>
            <p>The problem could be narrowed down to the plugin bcecause if you load the offending html into a texpane in a JFrame it works fine. Rather than waiting for an updated version to e fixed we came up with quite a simple solution, a solution that didn't take any effort at all. So what's it? we made a minor change to the CSS.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Java 5 bug and a work around

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I have always felt overture is a bit wierd and I recieved confirmation in the form of a cheque for the princely sum of $1.16.</p>
            <p>Since I live outside the US (in <a href="http://www.webquarry.com/~raditha/srilanka/">beautiful Sri Lanka</a>) it cost them $0.80 to mail it to me. I am sure they had to pay someone to print, sign and post it so it would have cost them much more. My bank charges $3.00 for each foreign cheque so if I deposit this into my account I end up losing $1.84 instead of gaining $1.16. I don't even know what the cheque is for!!</p>
            <p>This is not the first time I have seen a comedy of errors enacted by overture once upon a long ago I signed up to advertise through the overture channel but that story is worth it's own blog entry, so check back later. Suffice to say for the moment that google adsense is a far better option if you are a website owner and adwords is miles ahead of overture if you are an advertisee.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Overture and common cents

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>The site with the lamp has finally been '<a href="../../wiki/index.html">wikified</a>'</p>
            <p>If you are a frequent visitor here you might have noticed that I spent a bit of time looking at a few wiki scripts, a few days back I decided to <a href="529.html">end the quest</a> by settling on mediawiki as the chosen one.</p>
            <p>&nbsp;</p>
            <p>The chosen one is now at work on this. Though I wanted to hide away the edit pages behind a username and password ( you can do that by changing just a few lines of mediawiki code) I decided not to so so at this stage. When you are starting off a wiki, it's usually a blank other than for the system messages, you canno really expect people to contribute to such a blank wiki even if someone was inclined to make a contribution he is likely to run a mile when asked to register.</p>
            <p>&nbsp;</p>
            <p>So the wiki is editable by anyone, even with out a login until it is vandalised. If it ever becomes vandalised I will be pretty happy because that means this site is being noticed.</p>
            <p>&nbsp;</p>
            <p>update 2011-11-28 : The site was wikified but it kept getting vandalized till it could no longer be maintained. More about it <a href="bye-bye-wiki.html">here</a></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Wiki

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Just upgraded the j2se 5.0 installation on my computer from the beta to the full release. It seems the beta may have been more stable than the gold.</p>
            <p>Why do I say that? well after upgrading to jdl 1.5.0 (without the beta) the plug in always throws a hiccup with <a href="http://www.radinks.com/upload/">Rad Upload</a>. The very same applet that works on java version 1.4.1 to 1.5.0 beta throws quite a few nasty exceptions with the full release. I will append the stack trace to the end of this blog entry, you can check it out for yourself, you do not find any references to our code in the trace.</p>
            <p>So when exactly does the error occur? it happens just when the upload reaches completion and the contents of the drop area is about to be updated with the response returned by the server. I tried this with several versions of Rad Upload as well as several versions of java so I am pretty convinced that it's not our fault. Interestingly our offline tester (which is basically rad upload modified to run in a JFrame) works pretty well even with the latest version of java.</p>
            <p>Here is that stacktrace.:</p>
            <p>Exception in thread "AWT-EventQueue-13" java.lang.NullPointerException</p>
            <p>at javax.swing.text.html.CSS$FontSize.getValue(CSS.java672)</p>
            <p>at javax.swing.text.html.CSS$FontSize.toStyleConstants(CSS.java805)</p>
            <p>at javax.swing.text.html.StyleSheet$ViewAttributeSet.getAttribute(StyleSheet.java:2630)</p>
            <p>at javax.swing.text.StyleConstants.getFontSize(StyleConstants.java:358)</p>
            <p>at javax.swing.text.html.CSS$FontSize.getValue(CSS.java678)</p>
            <p>at javax.swing.text.html.CSS$FontSize.toStyleConstants(CSS.java805)</p>
            <p>at javax.swing.text.html.StyleSheet$ViewAttributeSet.getAttribute(StyleSheet.java:2630)</p>
            <p>at javax.swing.text.StyleConstants.getFontSize(StyleConstants.java:358)</p>
            <p>at javax.swing.text.html.CSS$FontSize.getValue(CSS.java678)</p>
            <p>at javax.swing.text.html.CSS$FontSize.toStyleConstants(CSS.java805)</p>
            <p>at javax.swing.text.html.StyleSheet$ViewAttributeSet.getAttribute(StyleSheet.java:2630)</p>
            <p>at javax.swing.text.StyleConstants.getFontSize(StyleConstants.java:358)</p>
            <p>at javax.swing.text.html.CSS$FontSize.getValue(CSS.java678)</p>
            <p>at javax.swing.text.html.CSS.getFontSize(CSS.java:614)</p>
            <p>at javax.swing.text.html.CSS.getFont(CSS.java:574)</p>
            <p>at javax.swing.text.html.StyleSheet.getFont(StyleSheet.java:867)</p>
            <p>at javax.swing.text.DefaultStyledDocument.getFont(DefaultStyledDocument.java:777)</p>
            <p>at javax.swing.text.LabelView.setPropertiesFromAttributes(LabelView.java28)</p>
            <p>at javax.swing.text.html.InlineView.setPropertiesFromAttributes(InlineView.java:229)</p>
            <p>at javax.swing.text.LabelView.sync(LabelView.java:39)</p>
            <p>at javax.swing.text.LabelView.getFont(LabelView.java89)</p>
            <p>at javax.swing.text.GlyphPainter1.sync(GlyphPainter1.java94)</p>
            <p>at javax.swing.text.GlyphPainter1.getSpan(GlyphPainter1.java:42)</p>
            <p>at javax.swing.text.GlyphView.getPreferredSpan(GlyphView.java:536)</p>
            <p>at javax.swing.text.FlowView$LogicalView.getPreferredSpan(FlowView.java:692)</p>
            <p>at javax.swing.text.FlowView.calculateMinorAxisRequirements(FlowView.java:216)</p>
            <p>at javax.swing.text.html.ParagraphView.calculateMinorAxisRequirements(ParagraphView.java28)</p>
            <p>at javax.swing.text.BoxView.checkRequests(BoxView.java:911)</p>
            <p>at javax.swing.text.BoxView.getMinimumSpan(BoxView.java:542)</p>
            <p>at javax.swing.text.html.ParagraphView.getMinimumSpan(ParagraphView.java:257)</p>
            <p>at javax.swing.text.BoxView.calculateMinorAxisRequirements(BoxView.java:879)</p>
            <p>at javax.swing.text.html.BlockView.calculateMinorAxisRequirements(BlockView.java29)</p>
            <p>at javax.swing.text.BoxView.checkRequests(BoxView.java:911)</p>
            <p>at javax.swing.text.BoxView.getMinimumSpan(BoxView.java:542)</p>
            <p>at javax.swing.text.html.BlockView.getMinimumSpan(BlockView.java:362)</p>
            <p>at javax.swing.text.html.TableView.checkSingleColumnCell(TableView.java:434)</p>
            <p>at javax.swing.text.html.TableView.calculateColumnRequirements(TableView.java:397)</p>
            <p>at javax.swing.text.html.TableView.calculateMinorAxisRequirements(TableView.java:519)</p>
            <p>at javax.swing.text.BoxView.checkRequests(BoxView.java:911)</p>
            <p>at javax.swing.text.BoxView.getMinimumSpan(BoxView.java:542)</p>
            <p>at javax.swing.text.BoxView.calculateMinorAxisRequirements(BoxView.java:879)</p>
            <p>at javax.swing.text.html.BlockView.calculateMinorAxisRequirements(BlockView.java29)</p>
            <p>at javax.swing.text.BoxView.checkRequests(BoxView.java:911)</p>
            <p>at javax.swing.text.BoxView.getMinimumSpan(BoxView.java:542)</p>
            <p>at javax.swing.text.html.BlockView.getMinimumSpan(BlockView.java:362)</p>
            <p>at javax.swing.text.BoxView.calculateMinorAxisRequirements(BoxView.java:879)</p>
            <p>at javax.swing.text.html.BlockView.calculateMinorAxisRequirements(BlockView.java29)</p>
            <p>at javax.swing.text.BoxView.checkRequests(BoxView.java:911)</p>
            <p>at javax.swing.text.BoxView.setSpanOnAxis(BoxView.java:325)</p>
            <p>at javax.swing.text.BoxView.layout(BoxView.java:682)</p>
            <p>at javax.swing.text.BoxView.setSize(BoxView.java:379)</p>
            <p>at javax.swing.plaf.basic.BasicTextUI$RootView.setSize(BasicTextUI.java631)</p>
            <p>at javax.swing.plaf.basic.BasicTextUI.getPreferredSize(BasicTextUI.java:825)</p>
            <p>at javax.swing.JComponent.getPreferredSize(JComponent.java582)</p>
            <p>at javax.swing.JEditorPane.getPreferredSize(JEditorPane.java227)</p>
            <p>at javax.swing.ScrollPaneLayout.layoutContainer(ScrollPaneLayout.java:769)</p>
            <p>at java.awt.Container.layout(Container.java401)</p>
            <p>at java.awt.Container.doLayout(Container.java390)</p>
            <p>at java.awt.Container.validateTree(Container.java473)</p>
            <p>at java.awt.Container.validate(Container.java448)</p>
            <p>at javax.swing.RepaintManager.validateInvalidComponents(RepaintManager.java:379)</p>
            <p>at javax.swing.SystemEventQueueUtilities$ComponentWorkRequest.run(SystemEventQueueUtilities.java13)</p>
            <p>at java.awt.event.InvocationEvent.dispatch(InvocationEvent.java:209)</p>
            <p>at java.awt.EventQueue.dispatchEvent(EventQueue.java:461)</p>
            <p>at java.awt.EventDispatchThread.pumpOneEventForHierarchy(EventDispatchThread.java:234)</p>
            <p>at java.awt.EventDispatchThread.pumpEventsForHierarchy(EventDispatchThread.java63)</p>
            <p>at java.awt.EventDispatchThread.pumpEvents(EventDispatchThread.java57)</p>
            <p>at java.awt.EventDispatchThread.pumpEvents(EventDispatchThread.java49)</p>
            <p>at java.awt.EventDispatchThread.run(EventDispatchThread.java10)</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Java 5 bug?

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="../../megaupload">Mega Upload</a> version 1.44 was released yesterday (7th Oct). There are two improvements over the previous version (1.43).</p>
            <p>There had been a few reports of the progress bar reaching 100% and not changing thereafter to the 'upload complete' state. I had seen this happen a few times myself but never really bothered about it. Fortunately a user of Mega upload from Germany, Nils Menrad did look into it and noticed that in some rare occaisions the browser seems to send more bytes than the Content-length header would indicate.</p>
            <p>The progress bar switches to the upload complete state when the number of bytes recieved equals the content length, this comparision would fail if the number of bytes exceeds the content-lenght. Of course it should never happen but apparently it does happen. So Nils suggested that the comparision should be equals or greater instead of equals. That's the way it is from now on in the 1.4 branch. The same change will be made to the 1.3 branch as well during the next few weeks.</p>
            <p>The second improvement over the previous version is that there is a replacement for progress.cgi in PHP. This contribution was made by Mike Hodgson. If you want to make any changes to the look of the progress bar you might find it easier to edit progress.php instead of progress.cgi if perl does not happen to be your favourite scripting language.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Mega upload 1.44

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Since the java class library that accompanies the JVM includes thousands of classes (and the number is growing all the time) it's nearly impossible to have ever used all these classes. I would be suprised if a developer who claims to have used every single class is ever found.</p>
            <p>Occaisionally when reffering to the apidoc, a class that has always been there but you never noticed suddenly catches your eye. That's what happened to me with javax.swing.filechooser.FileSystemView just now. Suddenly I was filled with the overwhelming temptation to write a few lines of code to figure out a practical use for this class.</p>
            <p>I knew that this class is being used by the FileChooser to create those icons but nothing more than that, I wanted to find out if the FileSystemView would be able to produce the same icon as konqueror or nautilus produces.</p>
            <p>Sadly it falls short. For me FileSystemView produced only two icons, one that symbolized a folder and another that symbolized a file. Fat use that is, I can easily do the same with just two three lines of code that includes a call to the File.isDirectory() method.</p>
            <p>There are some files in my home folder with wierd characters, some of them I occaisionaly use as bait on <a href="../../megaupload">mega Upload</a>, FileSystemView choked on them.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

FileSystemView

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>My <a href="525.html">quest for the perfect wiki</a> has finally ended. I have come a fully circle and finally decided to settle on mediawiki, the very product that I wanted to move away from.</p>
            <p>The mediawiki project is full of features but it's rather complex. The complexity arises from the fact that it is the software that powers the <a href="https://en.wikipedia.org">Wikipedia</a>, as a result it has to be many things to many people and perhaps not idealy suited for a small site like <a href="../../index.html">the site with the LAMP</a>.</p>
            <p>Then the Tavi project apears to be no less complex than the mediawiki but does not have as many features as mediawiki and it's very very slow.</p>
            <p>The phpwiki has a good set of features and it's certainly very flexible but one essential feature was missing. Since I do not have the time to keep an eye out against vandals I wanted to restrict the ability to edit the wiki to registered users. Users should be able to sign themselves up. Unfortunatley this part of the coding has not yet been completed in PHPwiki otherwise this is definitely what I would have chosen.</p>
            <p>I downloaded yet another wiki name wikiwig, unfortunately the documentation for this project turned out to be in a language I did not understand, so I could be sitting on a gold mine without knowing it.</p>
            <p>So after looking at just these three projects I have decided to stick with <a href="http://wikipedia.sourceforge.net">mediawiki</a> inspite of it's complexity and the fact that it uses mysql. After all there is a saying better the devil you know than the devil you don't. (I don't know the devil at all, haven't spent more than 10-15 hours on free BSD).</p>
            <p>If you have seen my writings on perl or Java you might ask why didn't I look at wiki software written in these languages. Since I started coding in PHP 3 many years ago, I have been reluctant to use perl for web application and I feel J2EE would be overkilll for something this simple and there is also the fact that the server this site is hosted on does not included a servlet container.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

The quest for the perfect wiki ends.

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="525.html">The quest for the perfect wiki</a> continues and today I take a look at <a href="https://sourceforge.net/projects/wikiwig/">wikiwig</a>. The first impression is negative because the download is an 'rar' file not a zip, gz or bz2. I managed to get it extracted by moving it to a windows now to try it out.</p>
            <p>Bad news the readme file is in french, a language I do not understand. Looked through the files and came up with what looks like the configuration file but the comments with in it are in french as well. Looks like wikiwig is going to make an early exit from my list. Now it's back to <a href="https://sourceforge.net/projects/phpwiki/">phpwiki</a>.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

what wiki to use?

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Yesterday we released the latest version of the <a href="http://www.radinks.net/ftp/applet/">FTP Applet</a>. This new release makes has some significant improvements over the previous versions.</p>
            <p>The most important change is the fact that you can now switch on and off between displaying hidden files. Hidden files as we all know are not too well hidden, in unix like operating systems all you need to do is pass an additional parameter to `ls` to display them. The behaviour of FTP servers isn't very different. The settings dialog can now be used to put on or take off the invisible cloak.</p>
            <p>We have had quite a few requests for the ability to delete non empty directories. We did not implement this feature upto now thinking that it would be too dangerous. Finally we decided that if that's what the clients want it's upto us to give it. So recursive directory deletion - or the ability to delete a folder that contains other files or subfolders is included in the 1.15 version of the FTP applet and deskto client. (Recursive folder upload and download has been supported for a long time)</p>
            <p>Both these features will soon find their way into our <a href="http://www.radinks.com/sftp/">SFTP Applet</a> as well.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

FTP Applet

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Have you read The Quest for Karla by John Le Carre? I have always felt that the pace of his books was too slow. My quest for the perfect wiki is not moving any faster.</p>
            <p>It was a week ago that <a href="520.html">I wrote about tavi</a> and it was more than an year ago that I tried out <a href="http://wikipedia.sourceforge.net/">mediawiki</a>. Today it's the turn of <a href="http://phpwiki.sourceforge.net/">PHPWiki</a></p>
            <p>phpWiki is a product that gives of good vibes. Not only is the product compatible with both <a href="https://www.postgresql.org">postgresql</a> and mysql but you can even use it without a database. You can store your data in flat files or even CVS.</p>
            <p>You would immidiately start thinking that the download size of 2MB is justifiable since this wiki gives you so many options. In actual fact most of the space on the tarball is taken up by themes - which means the code itself is very small; just the way it should be.</p>
            <p>When the wiki was loaded for the first time I ran into an error <em></em></p>
            <p><em>Fatal error: Maximum function nesting level of '64' reached, aborting! in /var/www/html/phpwiki/lib/BlockParser.php on line 280</em> Since this looks like an xdebug problem I edited the php.ini file to disable xdebug and gave it another go.</p>
            <p>I now have what the authors termed a 'virgin wiki' on my computer now to figure out how to get rid of the 150 debug messages that appear on the home page and to figure out how to get rid of this Fatal error:<em>Cannot instantiate non-existent class: file_passwd in /var/www/html/phpwiki/lib/WikiUserNew.php on line 2122</em>, the closest I found to a solution was at : <a href="https://sourceforge.net/tracker/index.html?func=detail&amp;aid=913911&amp;group_id=6121&amp;atid=106121">https://sourceforge.net/tracker/index.html?func=detail&amp;aid=913911&amp;group_id=6121&amp;atid=106121</a></p>
            <p>The problem was solved by modifying the WikiUserNew.php file along similar lines.</p>
            <p>Well the error message is no more but that does not mean I can login to the wiki. ( I do not want to allow anonymous edits) but enough for today.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

The quest for the perfect wiki

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>After downloading the tarball for version 2.6.8.1 version of the kernel and my trusty old compiler had a go at installing it. Other than the stability problems I mentioned <a href="000522.html">the other day</a> I have a couple of other reasons to upgrade the kernel.</p>
            <p>One is a recently aquired USB storage device. Sadly it's not been detected by my installation, if I have to compile the kernel again might as well make it an upto date version of the kernel. The second reason is that for some reason or other my creative webcam (which is nearly an year old) still refuses to capture anything. The closest I have come to video capture on linux is to make the light on the camera blink.</p>
            <p>While installing the last version I learnt a few things so there were no repeats of the <a href="000288.html">make+dep+is+unnecessary+now</a> error again. The reboot after the kernel was less than successfull it ended with the dreaded 'unable to mount root file system' error message.</p>
            <p>There are all kinds of wierd reasons while the root file system may not be mounted. In my case it turned out the the kernel was a slightly larger than the last time round and needed the help of a Ram Disk. Since I generally prefer to edit grub.conf by hand with out using the make install step the initrd had not been built.</p>
            <p>After booting up again in <a href="319.html">2.6.5</a> (followed by a lot of browsing that turned up nothing but dead ends) I did resort to the make install step and viola the I am now tapping away on a athlon running on 2.6.8</p>
            <p>The new kernel is still oblivious to the existence of the USB device (Creative Muvo TX by the way) and it looks like I may have to solve <a href="../../linux/rcalsa.php">the mystery of the dead speaker</a> all over again.</p>
            <p>Followup: oct 17 <a href="528.html">2.6.8.1 did solve the usb device problem</a></p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

linux 2.6.8

<p></p><p>The time has come to upgrade my kernel from what it is now (2.6.5) to the latest version (2.6.8). Since no one else logs into my machine I am not the slightest bit concerned about the vulnerability that was reported sometime ago with older kernels (I have forgotten the link, but a logged in user can crash your computer).</p><p>My real concern is that the machine appears somewhat unstable. During the past two weeks it has restarted for no apparent reason. Initially I suspected the newly installed RAM to be the culprit, it was and it wasn't. The stick had not been placed very tightly so it probably was the culprit on a couple of occaisions, but it was removed and plugged in again with much more care and now the RAM would be aggrieved if I blamed the crashes on it.</p><p>Part of the problem could be my <a target="_blank" rel="noopener noreferrer nofollow" href="507.html">botched gnome upgrade</a> attempt it has without doubt lead to quite a few compatibility issues but that shouldn't cause the desktop to freeze up when I try to set up a PPP connection over GPRS.</p><p>Curiosly once when an involuntary restart happened the start up scripts refused to go past the fsck stage. It would check half a partition and then restart. I managed to overcome this by booting into my java desktop installation and running fsck from there. I can't rule out the possibility of some important libraries or binaries being corrupted. I did lose my <em>daemontools</em> and qmail installations because of a restart that left some dangling inodes.</p><p>All this points to the fact that something is not right and the easy way out is try to recompile the kernel let's see how it goes.</p><p></p><p></p>

Time for a kernel upgrade

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>It's been an year since I first tried out the mediawiki project. Since then I have modified/customized several versions of the software but I have not aquired a taste for it yet. The reason is that <a href="http://wikipedia.sourceforge.net/">mediawiki</a> is extremely complicated. Extremely complicated does not mean bad. This is a very good piece of software, it handles the gigantic database of the wikipedia project extremely well. It's just that I don't like complex projects that have too many features.</p>
            <p>In recent times there have been some discussions in 'the list' about wikis and one of the names that came up was <a href="http://tavi.sourceforge.net/">tavi</a> aka WikkiTikkiTavi. I decided to give this a try and i really liked the size of the download the same cannot be said about the installation instructions. It's amazing that such a small piece of software has so many installation steps.</p>
            <p>The configure script asks you a series of questions before producting the configuration file. This reminds me of the bad old days of when you didn't have the luxury of using make xconfig when compiling the kernel. This task certainly feels more complex than building the kernel.</p>
            <p>Having put my nose to the grinder my reward was an error message:</p>
            <p><code>Error executing database query.</code></p>
            <p><code>Please contact the administrator for assistance.</code></p>
            <p>LOCK TABLES rate WRITE</p>
            <p>Not sure what the cause of the error is but you find lot's of wierd suggestions at : <a href="http://tavi.sourceforge.net/TaviNoPerl">http://tavi.sourceforge.net/TaviNoPerl</a>. What worked for me was to go into the admin section click lock/unlock pages and then save. That apparently created the first page in the wiki and the error message disappered.</p>
            <p>Boy is this slow or what. The page takes up quite a few seconds to load with just one record in the database. I shudder to think what might happen with hundreds of pages. I first thought the authors had gone overboard with table locking but soon I found the real reason.</p>
            <p>Each function with in the tavi scripts attempts to establish a database connection that is a huge huge overhead. Now you might think using a persistent connection will eliminate that overhead. It does not. Even if you use a persistent connection the pool has to be queried and if an old connection is reused you still need to test if it's alive what a waste of clock cycles.</p>
            <p>The reason that I am looking for alternatives to mediawiki is because I want something less complex, tavi does not seem to be the solution.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Making a Wiki

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Despite years of programming I have never quite got the hang of date time functions. They all look pretty similar and cumbersome to use regardless of programming language or RDBMS.</p>
            <p>Date and time functions in mysql are particularly confusing though less confusing than the counterparts in PHP. To simplify things I have been in the habit of calling the unix_timestamp() function to covert evertying to longs, according to what I have seen I am not the only one to use this bad pracitice.</p>
            <p>Unfortunatley unix_timestamp() is not supported in <a href="https://www.postgresql.org">Postgres</a>. I did come across a suggestion that seems to work pretty well: <em>select round(date_part('epoch', my_date)) </em> the compatibility problems do not end there. There are several other incompatibilities between the mysql and postgres and I intend to take a closer look at this during the coming weeks.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

pgsql, mysql and date time functions

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Due to to a bungling by the bank that issued my credit card my AITCOM woes continue. A few weeks back I wrote on <a href="498.html">how to cancel an AITCOM account</a>. The only way seems to be to cancel your credit card. That is indeed what quite a few others seem to have done as well.</p>
            <p>Unfortunately, Cancelling your credit card does not mean you automatically get a refund for the <a href="491.html">money they have wrongly taken from you</a>. For that you have to dispute the transaction and ask for a chargeback, which I did. The bank offered me a temporary credit 'pending investigation'.</p>
            <p>After nearly two months I recieved a call from the bank saying they have had to reverse the chargeback because ait has provided documentary evidence about the transaction.</p>
            <p>According to my understanding this documentary evidence has to be letters on paper with a signature , or at least that's what <a href="https://www.2checkout.com/">2checkout</a> the CC service provider we use tells us. AITCOM definitely does not have such signed documents, so i assume that my bank (which still lives in the dark ages) has finally started to accept electronic documents as proof. So I forwarded some emails and staked a dispute agains the resolution of the chargeback. let's see what comes of it.</p>
            <p>In the meantime read some <a href="446.html">more about AITCOM</a></p>
            <p><a href="http://poza.net/ait/">http://poza.net/ait/</a></p>
            <p><a href="http://www.aitsucks.com/cgi-bin/yabb/YaBB.cgi?board=news;action=display;num=1092947306">http://www.aitsucks.com/cgi-bin/yabb/YaBB.cgi?board=news;action=display;num=1092947306</a></p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

AITCOM woes continue

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="../../megaupload">Mega Upload</a> has been among the top 1% of projects listed at <a href="https://www.sf.net/">sourceforge</a> for0 days running. If I remember correctly this is the longest period during which it has been in the top 1%</p>
            <p>During this perioud it reached a high of the 38th most active project quite something for a download that's only about 20kb. It used to be smaller, less than 10kb right at the start but as more and more improvements were added it has grown to be double what it was. That's still about 1/100 the size of other active projects at sourceforge.</p>
            <p>I have been putting off the next release for a quite some time, I guess it is now time that I did so.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

10 days at the top

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>When I tried out the <a href="512.html">Sinhala Knoppix</a> a few days ago, that was the first time I had seen a Ram Disk in action since the days of MS DOS.</p>
            <p>Well to be precise, I should have said a Ram Disk mounted after the operating system had booted up. It's quite common for linux to make use of a Ram Disk at boot time, in fact I have a java desktop installation on my computer that does just that.</p>
            <p>I just bought another stick of RAM yesterday and that made me quite keen to try to mount a Ram Disk. I know that linux would cache files in memory when extra memory is available and speed things up, but that speed up applies only for read operations and not for write operations. Compiling software involves a lot of writes.</p>
            <p>Mounting the Ram Disk turned out to be quite a trivial tasks. I did find a couple of interesting articles at <a href="http://www.aleph1.co.uk/armlinux/book/mkramdisk.html">http://www.aleph1.co.uk/armlinux/book/mkramdisk.html</a> and <a href="http://www.lissot.net/partition/ramdisk.html">http://www.lissot.net/partition/ramdisk.html</a></p>
            <p>Unfortunatley after using it for a few hours, it started to produce IO errors so I decided to give up on RAM disks since there wasn't such a great improvement in speed after all.</p>
            <p>Follow up: <a href="000527.html">Ramdisk vs tempfs</a></p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Ram Disk on linux

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>The java version of <a href="../../megaupload">megaupload</a> has finally seen a new release after a long long time. It's entirely thanks to a code contribution made by one of it's users.</p>
            <p>Simon Chiaretta liked the <a href="../../megaupload/jsp.php">JSP edition</a> so much that he spent some time on it to bring it upto a level that is comparable to the other editions. As a result of his efforts the commons edition of megaupload now displays the time elapsed and upload speed just the same way that the PHP or perl editions do.</p>
            <p>On a side note, megaupload is once again in the top 1% at <a href="https://sourceforge.net/projects/megaupload">sourceforge</a>. Interestingly the project made it's way into the top 1% before the release was made. Usually you tend to see a project leap frog other popular projects immidiately after a new version has been released. This time around the movement came even before the new release.</p>
            <p>What's special about mega upload when compared to other projects is that it's such a small download. In other words it's a really simple solution to what appeared to be complext problem.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Mega Upload - Finally a new JSP edition

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="44.html">It was in march that I downloaded java help</a>. It took just one look for the realization to dawn that the help on java help needs a lot of help. So it has been in the back burner for six months. Today I had another look at it and this time struggled through a few chapters of extremely boring documentation before I figured out how to run the demos.</p>
            <p>In most software packages demos are supposed to work with minimal effort and can usually be executed without reading through the manual. Not so java help. You might think authors of a help system might know better.</p>
            <p>To be precise how to run the demos wasn't part of the manual, it was hidden away in one of the numorous README files scattered throughout the download bundles. In order to run the demo you need to compile a few source files. The class paths, compiler directives etc are all spread across a page and a half of this readme and becomes quite a tedious cut and paste operation.</p>
            <p>One wonders why compiling the demos isn't an ant task. In fact there is an ant task to compile the so called server help and this has to be compiled separately from other demos. The developers of Java Help certainly have done their level best to make the system extremely annoying and dfficult to use.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Java Help

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>On the surface handling file uploads with PHP is a piece of cake (then why do we see so many messages in mailing lists on this topic?). It's what's under the hood that causes the problem.</p>
            <p>For example most installations of PHP are geared only to handle a maximum total file size of 2MB. Many users who are on shared hosting enviorenments cannot edit their php.ini file so they are unable to move past this limit. That's why I wrote that article '<a href="../../megaupload/upload.php">Sneak Past the PHP upload limits</a>'</p>
            <p>PHP 4.3.8 and 5.0+ adds another complication, this time it's not related to upload limits but to path information. In previous version PHP discarded all pathnames that appeared to be windows paths. (in other words 'c:downloadsraduploadlite.zip') would appear as 'raduploadlite.zip' but '/downloads/raduploadlite.zip' would be left unchanged.</p>
            <p>We made use of this feature (it's a feature not a bug!?) of PHP in <a href="http://www.radupload.com/">Rad Upload</a>. Rad Upload can translate pathname by replacing the windows path separator with the linux path separator so that PHP does not discard the full path information.</p>
            <p>Now apparently the authors of PHP believe pathnames should be completely stripped out. A good programming language should give the developer enough rope to be free or to hang himself. Apparently now PHP developers are cannot be trusted not to wrap wrap the rope around their necks and strangle themselves.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

PHP 4.3.8 and file uploads

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I made a startling discovery that could potentially make the mysql jdbc driver obsolete. More on that later I need to investigate it some more. In the meantime I decided to catch up with what others are saying about the license issue. Curiously there seem to be a shortage of new entries most of what I found were posted at least before July.</p>
            <p>the blog; <a href="http://www.schlossnagle.org/~george/blog/archives/236_MySQL_Licensing_Continued.html">a dash of technology and a pinch of life</a>; tells a story that's pretty much like my own encounter with mysql sales representatives. He had been offered a client license, in my case the sales rep didn't know about the existense of such a license and wanted us to by a license for a server that we will never redistribute.</p>
            <p><a href="http://blog.megacity.org/archives/001657.php">Derek Balling</a> has done a very good job of taking the wind out of some of the wacky statements made by mysql AB.</p>
            <p><a href="http://jeremy.linuxquestions.org/blog/_archives/2004/8/16/124876.html">This blogger</a> on the other hand is still a firm believer in the mysql gospel.</p>
            <p><a href="http://www.lethargy.org/~jesus/archives/23_MySQL_the_best_database_Im_afraid_to_use.html">Another blog by Theo Schlossnagle </a> is pretty close to my own attitude to paying mysql license fees as well.</p>
            <p>Having read these entries and many others that you will run into in a countless mailing list archive you will probably want to give up on mysql all together and move to <a href="https://www.postgresql.org">Postgresql</a>. If you do, hope you find my <a href="../../mysql/mysql2pgsql.php">mysql2pgsql</a> schema migration tool helpfull.</p>
            <p>Finally a sense of fair play says I should <a href="http://www.phpmag.net/itr/online_artikel/psecom,id,618,nodeid,114.html">link to mysql AB's version</a> of the story as well but please do keep your salt shaker at hand.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Mysql License

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Today I picked up a <a href="http://sinhala.linux.lk">Sinhala Linux</a> CD at the Colombo Internation book fair. The book fair itself is an event worth talking about but more on that later. The CD is a modified version of the <a href="http://www.knoppix.org">knoppix</a> worked on by a bunch of Sri Lankan geeks.</p>
            <p>These geeks belong to two non profit organizations, one being the <a href="http://tux.metta.lk/">MettaNet Foundation Tux Group</a> and the other being the <a href="http://www.linux.lk">Lanka LUG</a>. Neither organization has money to spare so the stall wasn't eye catching but according to what I gathered a steady stream of customers have been visiting to buy the cd priced at Less than US$ 0.50</p>
            <p>Apprently the demand was so high that they later had to double the price. In my opinion they should have priced it much higher so that some funds could have been generated for the LUG's activity the $0.50 barely covers the cost of the media.</p>
            <p>Unfortunately for me knoppix has never worked on my computer because the MSI VGA card on it doesn't get detected, the same fate befell the sinhala version but I did have success running it on another computer with a ATI card.</p>
            <p>The demo at the book fair even had the menu in Sinhala but unfortunately the CD that has been sold does not have a sinhala menu but still it is indeed a comendable effort.</p>
            <p>Getting on to the book fair it looked as if all of colombo's 1 million population was crammed into the exibition venue. No doubt by the time the exibition ends a good percentage of the country's population would have seen it.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Sinhala Linux

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>The time has come to upgrade my <a href="https://www.postgresql.org">postgres</a> installation from the ancient 7.3.2 currently running on my computer to the more recent 7.4.5 version.</p>
            <p>Upgrading is a bit tricky because the pg_dumpall function in the older versions isn't as good as it could be. The developers have presented a chicken and egg solution for this problem: you are supposed to run the pg_dumpall function of the <em>version you are upgrading to</em>.</p>
            <p>A good installer will not clobber your old data but the mantra of software development is back up, back up and back up again. So I took a copy of my <em>/var/lib/pgsql/</em> directory just in case. If you are on Red Hat this is the location you are likely to find your data. (btw my system is nominally redhat but only a few Red Hat RPMs are left on - postgresql is one of them.)</p>
            <p>Since I will be accessing my database from java as well as PHP my configure command looks like:</p>
            <p><code style="color: #003366">./configure --datadir=/var/lib/pgsql/data --with-java</code></p>
            <p>This step didn't produce any suprises so onto the next step. The docs say use <em>gmake</em> which is the same as <em>make</em> on my system, so here goes ....</p>
            <p>It did take a while to complete the compilation (enough time for me to add a few lines of code to a nother project I am working on) and it wasn't successfully. The break was when attempting to copmile the jdbc driver. I already have a more upto date version of the driver which I downloaded a couple of weeks back as a binary. So here goes another attempt to compile postgres, this time without the jdbc driver. (time to add a few more lines of code but not too many this time).</p>
            <p>The second run was uneventfull. But before going on to the <em>make install</em> step I now need to run the 'new' pg_dumpall command to extract the data from my database. Too bad it does not work without the make install step.</p>
            <p><code style="color: #003366"></code></p>
            <p><code style="color: #003366">./src/bin/pg_dump/pg_dumpall: relocation error: ./src/bin/pg_dump/pg_dumpall: undefined symbol: get_progname</code></p>
            <p></p>
            <p>Hang on a sec, what if I remove my RPMs? no luck. So now it's time to say bye bye to the data (I don't need it anyway, I already have the schema's safe and sound on numorous CDs) and on to the <em>make install</em></p>
            <p>Make install went well but attempting to run initdb causes problems. Apparently I am not supposed use --datadir=/var/lib/pgsql/data because the error produced says:</p>
            <p><code style="color: #003366">The database cluster will be initialized with locale en_US.UTF-8.</code></p>
            <p><code style="color: #003366">initdb: directory "/var/lib/pgsql/data/" exists but is not empty</code></p>
            <p><code style="color: #003366">If you want to create a new database system, either remove or empty</code></p>
            <p><code style="color: #003366">the directory "/var/lib/pgsql/data/" or run initdb with an argument other than</code></p>
            <p><code style="color: #003366">"/var/lib/pgsql/data/"</code></p>
            <p></p>
            <p>Attempting to clear out the folder leads to another error:</p>
            <p><code style="color: #003366"></code></p>
            <p><code style="color: #003366">The database cluster will be initialized with locale en_US.UTF-8.</code></p>
            <p><code style="color: #003366">initdb: directory "/var/lib/pgsql/data/" exists but is not empty</code></p>
            <p><code style="color: #003366">If you want to create a new database system, either remove or empty</code></p>
            <p><code style="color: #003366">the directory "/var/lib/pgsql/data/" or run initdb with an argument other than</code></p>
            <p><code style="color: #003366">/var/lib/pgsql/data/</code></p>
            <p></p>
            <p>So it's time for another compilations. This time my configure script has no paramaters. (This entry will probably end up being the longest in my blog). Thankfully this time everything including the initdb went well.</p>
            <p>My problems do not end here. I now need to recompile my php installation. Since I have both PHP 4 and PHP 5 running (no not together, i have to edit httpd.conf to use one or the other) that means a lot more work for my poor old hard disk and CPU.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Postgresql Upgrade

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Another victim of an upgrade on my system is <a href="https://jakarta.apache.org">Tomcat</a>. Though a 5.5 version has been released I decided to upgrade from the current version is 5.0.18 to 5.0.28 instead of using the more recent version.</p>
            <p>It's been rare for tomcat to display major differences in behaviour between minor version numbers. However code that executes perfectly well in 5.0.18 refuses to run on anything more recent than version 5.0.25.</p>
            <p>Besides all that the <a href="http://www.printingdemo.com">online demo</a> for the <a href="http://www.radinks.com/print/">print suite</a> uses a more recent version of tomcat and it certainly pays to have the installation on my computer in sync with what is on the server.</p>
            <p>My next victim is going to be either <a href="https://www.postgresql.org">postgresql</a> or the <a href="http://java.sun.com">j2sdk</a>. The postgresql version I have installed is 7.3.2 whereas the current version looks to be 7.3.5</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Tomcat Upgrade

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>The <a href="http://www.radinks.com/print/">Print Suite</a> has barely hit the shelves but we have a sale already. So much for the saying that IT products of this nature take at least 60-90 days to start moving.</p>
            <p>The print suite was <a href="505.html">finally released on the 31st</a> after a long delay so it's taken just four days for the sale to occur. This is in spite of the fact that we had to keep working on the <a href="http://www.printingdemo.com/">demo</a> for this product even after the release. That meant the demo occaisionally threw up errors or had other failing as the servlet container was reloaded or restarted.</p>
            <p>All is well that ends well as we have been able to chalk up a sale. We had similar pleasent experiences with <a href="http://www.radupload.com/">Rad Upload</a> and <a href="http://www.radinks.com/sftp/">Rad SFTP</a>. Let's hope we can notch up a few more sales in the coming weeks.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Sale

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Yesterday's story was about a Jmeter installation that dragged over many moons. Here is another. I began an <a href="503.html">installation of gnome 2.6.1 from source</a> a quite some time ago.</p>
            <p>Don't for a moment think I spent six weeks doing nothing but compile gnome, instead I would multitask running the configure, make , make install in the background while doing other work in a foreground process. Naturally thread priority was for the foreground task, some days the background task never had even as much as a clock tick</p>
            <p>Anyway in the end finally reached the stage where gnome would (reluctantly) start up but as the screen shots will reveal it's a very reluctant installation indeed.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Gnome Upgrade

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Many moons ago <a href="390.html">I found a tarball of JMeter</a> in my downloads folder. Finally I decided to have a go at installing it a task that turned out to be rather more complicated than it should have been.</p>
            <p>To start with; <a href="https://jakarta.apache.org">Jmeter source</a> that I downloaded does not seem to be compatible with java version 1.5 (<a href="000437.html">or 5.0</a>). Ant exits with the following errors.</p>
            <p><code style="color: #003366; margin:15px"></code></p>
            <p><code style="color: #003366; margin:15px">[javac] javac: target release 1.2 conflicts with default source release 1.5</code></p>
            <p><code style="color: #003366; margin:15px">BUILD FAILED</code></p>
            <p><code style="color: #003366; margin:15px">/opt/java/jakarta-jmeter-2.0.1/build.xml:633: Compile failed; see the compiler error output for details.</code></p>
            <p></p>
            <p>That was easily fixed by changing the JAVA_HOME and the PATH so that ant made user of the older.4.2 installation than the 1.5 version. That resulted in the Jmeter binary being compiled but the documentation wasn't ready yet. As an added curiosity, once compiled jmeter would happily start up even if the JAVA_HOME was set to the 1.5 installation.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

JMeter

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>At <a href="http://www.radinks.com/">Rad Inks</a> we have finally been able to release our long delayed <a href="http://www.radinks.com/print/">web based printing software</a>. The reasons for the delays are numorous but one of the major causes is that we have been too distracted with our other products and working on customizations.</p>
            <p>Another major stumbling block that we ran into was the licensing issues of the mysql database. Initially we planned to make print suite compatible with mysql and microsoft sql databases (microsoft is not my choice). Then mysql decided to interpret the GPL in completely different way from everyone else. They caused a lot of uncertainty about whether or not our clients would need to pay license fees. We are not the only company to be faced with this uncertainty but our approach was to dump mysql all together and move towards postgresql.</p>
            <p>Well we have finally managed to overcome all obstacle and as a result the print suite became available for download for the first time on 31st of August.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Print Suite released (at last)

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="503.html">My previous blog entry</a> was made after I had reinstalled some of the older (v1.x) gnome RPMs and cajoled mozilla into starting up. I continued with the compilation of the new version while browsing. In the middle of all that the browser suddenly browser disappeared from the desktop. Attempts to restart would result in the following error</p>
            <p><em>Gtk-ERROR+**: Unable to load default font</em></p>
            <p>Unfortunately google wasn't much help so I decided to look for more suspect gnome libraries and reinstall them. The prime suspect was <em>gnome-libs-1.4.1.2.90-32.i386.rpm</em> but it had a few dependencies so I ended up doing.</p>
            <p><code>rpm -i RedHat/RPMS/gnome-libs-1.4.1.2.90-32.i386.rpm</code></p>
            <p><code>rpm -i RedHat/RPMS/ORBit-0.5.17-7.i386.rpm</code></p>
            <p><code>rpm -i RedHat/RPMS/GConf-1.0.9-10.i386.rpm</code></p>
            <p><code>rpm -i RedHat/RPMS/gnome-vfs-1.0.5-13.i386.rpm</code></p>
            <p></p>
            <p>And the reluctant mozilla (mozilla 1.7) to be exact agreed to start up again. It's such a relief not having to browse with Konqueror. Though konqueror is my default file manager, it's not the best web browser available for the linux desktop.</p>
            <p>And while all this was happening, version 2.6 installation came to a halt with the following error.</p>
            <p><code></code></p>
            <p><code>/bin/sed: can't read /usr/lib/liblinc.la: No such file or directory</code></p>
            <p><code>libtool: link: `/usr/lib/liblinc.la' is not a valid libtool archive</code></p>
            <p><code>make[3]: *** [libthemus-theme-properties-view.la] Error</code></p>
            <p><code>make[3]: Leaving directory `/opt/gnomesrc/control-center-2.6.1/vfs-methods/themus'</code></p>
            <p><code>make[2]: *** [all-recursive] Error</code></p>
            <p><code>make[2]: Leaving directory `/opt/gnomesrc/control-center-2.6.1/vfs-methods'</code></p>
            <p><code>make[1]: *** [all-recursive] Error</code></p>
            <p><code>make[1]: Leaving directory `/opt/gnomesrc/control-center-2.6.1'</code></p>
            <p><code>make: *** [all-recursive-am] Error 2</code></p>
            <p></p>
            <p>God knows why it's looking for an la but I managed to dig it up in the ORBit2-2.10.2 source tree - obviously the make install step had not copied it to the lib folder. After doing that manually the compilation of the control-center, which gave the hiccup went to it's completion.</p>
            <p>The same cannot be said about the make install step. It complained about a missing '.libs/libthemus-theme-properties-view.lai' this is something that's supposed to have been created in the <em>make</em> but it hasn't happened. Closer inspection revealed that a missing <em>intltool</em> tool might be the culprit here so I installed 'intltool-0.30' and managed to clear up the problem.</p>
            <p>That wasn't the end of it, the liblinc error appeared again I finally managed to bury it for good by changing into the <em>ORBit2-2.10.2/linc2</em> directory and hitting make and make install.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Gnome and Mozilla

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>As this blog entry is been written the CPU is quitely ticking away while the hard disk whirls at the task of compiling gnome 2.6.1 from source. You will no doubt wonder why I am installing 2.6.1 now when 2.6.2 is already available. Well I downloaded this 2.6.1 version some time back and as usually happens with projects like this it never got off the ground.</p>
            <p>I am simply too lazy (and too stingy - downloading is as expensive as driving a BMW) to bother downloading the new version of a desktop that i will rarely use (I work almost excluseively on KDE) Which version? 3.1 would you believe!</p>
            <p>Installing from source code is not for the faint hearted. There are numerous packages and unless you install them in the right order you get nothing more than a heap of errors about failed dependencies.</p>
            <p>It took me quite a while to figure out this sequence. Unfortunately i didn't write it down so if I have to install gnome again I have to go through the whole painfull process again.</p>
            <p>While bogged down with this I decided to uninstall the existing (already badly damaged) version of gnome on my machine. This was a big mistake; not just because the uninstallation is painfull affair but because lots of software still depend on version 1.2 of glib.</p>
            <p>One of the most prominent products that depend on this older version of glib is mozilla, I had to google with konqueror to figure out that the only solution is to reinstall the glib-1.2 which I did from RPMs.</p>
            <p>Right-O, I have come to the end of the blog enty but gcc is still hard at work, will let you know about the outcome tommorow.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Gnome 2.6.1 upgrade

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>If you have been a regular visitor to this site you may be wondering about the remarkable lack of posts during this month. It's been a week since I last made a blog entry. What is the reason? there are two reasons in fact.</p>
            <p>The first reason is that it has been a crazy couple of weeks at <a href="http://www.radinks.com">Rad Inks</a> with the <a href="500.html">servers failing with alarming regularity</a>. To comphensate for that we resorted to DNS round robin but that lead to more headaches than it solved - at least temporarily.</p>
            <p>The reason for the headaches is that our servers are not identical mirror images of each other. Being a small company we simply cannot afford that sort of redundancy in hardware. What we did was to draft in a server that was doing other duties to function as a web server and it lead to hard to find configuration errors which we are still resolving.</p>
            <p>One such error was that the second server was acting as the demo for the <a href="http://www.radinks.com/print/">print suite</a>. As such any request on any virtual host for the print/ folder would be routed to the demo. Ouch!</p>
            <p>So what's the second reason for not blogging as often? well having to spend most of my time on troubleshooting has meant that I haven't had the time to do any research of late. Hope to get started on that over the weekend.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Tapering off?

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>August 11th was the first birthday for <a href="../../megaupload">Mega Upload</a> the birthday coincided with the latest round of network problems so I was in no position to celebrate - to make matters a little worse the demo and the site for megaupload was largely inaccessible yesterday. Hosting for this site is provided by <a href="http://www.radinks.com/">Rad Inks</a> and thus the <a href="500.html">networks woes</a> were felt here as well.</p>
            <p>Anyway, things are back to normal though I am still worried about the demo. The reason is that like <a href="http://www.radinks.com/">RadInks.com</a>, this site is now load balanced by way of DNS round robin. Though my own testing has not revealed any problems with the demo I am still worried about how it might behave for some users, particularly those browsing from behind proxies.</p>
            <p>The birthday also gives more reasons to get the next release out and perhaps start off on the C edition which I have been threatening to do for a long time. In the meantime the three editions currently available for download have seen around 38,500 downloads over the year. Not bad for a project that's still less than 25kb.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Happy Birthday Megaupload

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="http://www.radinks.com/">Rad Inks</a> is having growing pains. Until today we <a href="http://www.radinks.com/">radinks.com</a> was hosted on just a single server. This server has been crashing on an all too often of late.</p>
            <p>The worst outage occured today and yesterday all in all the server was inaccessible for around 14 hours out of the last 48 - mails began to bounce and sales screeched to a halt. It was then that we decided to a do a load balancing act.</p>
            <p>We already had another server with us that wasn't under heavy load but the problem was with the DNS - not having a residential DNS expert we used the DNS servers of <a href="http://www.webquarry.com/">our ISP</a> upto now we switched those across to <a href="https://www.zoneedit.com">zoneedit</a> which offers excellent domain names services for a very reasonable fee.</p>
            <p>You can easily set up a round robin system and what's more zoneedit can detect when one of the servers are failing and temporarily remove that IP from the list. In other words you get a very simple fail over system for a very very reasonable price.</p>
            <p></p>

        </div>
    </div>


</body>

</html>

Growing Pains

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>If you have done any object oriented programming with PHP 4, you may have been left with the impression that it was an after thought. PHP 5 on the other hand includes several major improvements to the way that objects are handled. But there is a but. They appear to have been torn by a dilemma, to be or not to be like java.</p>
            <p>In the older version Constructors used to have the same name as the classname - exactly the way it is in Java. You can still name your constructors in the same way but now the recommended approach is to name the constructor as <em>__constructor</em> which is kind of ugly.</p>
            <p>Destructors are non existent in Java but they are part of the landscape in C++ and they are now a part of the PHP5 as well. Still it makes you wonder why destructors should be named as <em>__destructor</em> instead of with the class name prefixed with <em>~</em> as done in C++.</p>
            <p>Another java like improvement is the <em>__clone</em> method, i have still not been able to determine if this is a shallow copy or a deep copy but as in java you can easily override this method to make it behave the way you want it to behave.</p>
            <p>Other new kids on the block include abstract classes and method both of which seem to behave similar to their counterparts in Java and the same can be said of Intefaces.</p>
            <p>How about parameter passing? in Java you will find that primitives are passed by value and objects are passed by reference, in PHP 4 everything was passed by value unless you explicitly passed by reference. Of course explicitly passing by reference is not something that you could do with java. The approach in PHP 5 is to mix the best of both worlds. Parameter passing now mimics the behaviour of java but you can use the good old ampersand when you want to.</p>
            <p>We now have with us a group of classes that are 'built in' or a standard set of classes. Someone unfamiliar with PHP might not be impressed by their small number, but the strength of PHP lies in it's massive collection of functions which can do pretty much the same thing that java's 'inbuilt' set of classes can do. Some say more easily than with java.</p>
            <p>The most notable built in class is the Exception class. Yes try - catch is now a part of PHP. The throw keyword is also available but I have not yet dug up any references to the throws keyword.</p>
            <p>While still on the subject of keywords the private and protected keywords are also a part of PHP5, but because PHP 5 does not seem to have packages the so called default modifier (which is the absence of a modifier) is missing. In the absence of a modifer in php the class, method or property becomes public while in java it becomes what is commonly known as 'package private'. Interestinly I have not seen mention of the public keyword in the php docs.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Object Oriented PHP vs java

<p></p><p>Many developers and DBAs who spent their lives on M$ SQL or Oracle turn up their noses at mysql, they refer to it with disdain. According to them the lack of foreign key support in mysql means it is not a proper RBMS and not standard compliant. Nothing can be further from the truth.</p><p>Mysql has supported foreign keys for a long time now many people simply don't know of it or use it because foriegn keys cannot be used with <em>MyISAM</em> type of tables. You need to use <a target="_blank" rel="noopener noreferrer nofollow" href="http://www.innodb.com">InnoDB</a> type tables instead.</p><p>Since many of the mysql tutorials on the web were written long long ago before InnoDB support was added, they rarely refer to the use of this table type or foreign keys. Fortunately it's a trival matter to convert your tables from MyISAM to InnoDB - provided that their data integrity has not been violated.</p><p><em>ALTER TABLE tablename TYPE=InnoDB</em> is all that is needed to perform the conversion. Now you are ready to add foreign keys - let's look at an example.</p><p><code>alter table userProfile add Foreign key (userId) REFERENCES users(userId);</code></p><p>This foreign key is useful as it can stop you entering a User Profile for a non existent user. Even more useful is to have a foreign key that results in an unwanted user's profile being deleted when you delete the entry in the user's table. You an also prevent a use from being deleted at all if another table refers to that user. Before we create such a foreign key,  let's drop the old one we created. </p><p><code>alter table userProfile drop foreign key userId;</code></p><p>Beware; this drop foreign key statement will fail if there are inconsistencies in your data. The only remedy then is to create a new table do an INSERT .. SELECT query , drop the original table and rename the new table to use the old table name then take a deep breath.</p><p>Before we were so rudely disturbed, we were discussing cascading deletes, here is how you should create your foreign key if you want to make use of this feature: </p><p><code>alter table userProfile add constraint foreign key (userId) references users(userId) on delete cascade;</code></p><p></p><p></p>

Foreign Keys and mysql

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>As the name implies <a href="http://www.radinks.net/user/">Rad User Manager</a> is a login and user management systems for websites. This product was developed from the PHP scripts that power the <a href="http://www.radinks.com/members/">Rad Inks members area</a></p>
            <p>Just yesterday we decided to change to license model for Rad User - in other words the product is now open source and free. Since it's covered under the Mozilla Public License you can even integrate it with commercial products - as long as full credit is given. That usually means you have to mention in your readme file or release notes that the product makes use of Rad User Manager.</p>
            <p>Now you may ask isn't all this unfair on the people who paid for the license? well we had a survey running for quite sometime on our site and people who took part in it were entitled to download the User Manager free of charge - as a result we ended up not selling any licenses at all.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Rad User Manager Open Sourced

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>A couple of days ago I put up a simple <a href="../../mysql/mysql2pgsql.php">script</a> that can convert from mysql table structures to <a href="https://www.postgresql.org">postgresql</a>. Currently it does not work with the actual data but can give you a script that you can run on your postgres database to recreate the same set of tables that you find in your mysql db.</p>
            <p>As for the importing the data itself, I am currently doing something about that as part of the <a href="http://www.radinks.net/barbet/">Barbet</a> project. Barbet is a database admin tool that I am working on for <a href="http://www.radinks.com/">Rad Inks</a></p>
            <p>It is far simpler to migrate the data using an updatable ResultSet than to use PHP or to use insert queries. The script that I have described above utilizes regular expressions (no doubt the purists are frowning) to convert the sql script from mysqldump to one that is compatible with postgres.</p>
            <p>Doing the same with the actual data is more difficult because of it's much harder to predict what it may be. A s a result the insert statement would have to deal with quite a bit of type conversion and escaping special characters.</p>
            <p>Updatable ResultSets on the other hand can be used to insert data without the use of an SQL query - well query does happen behind the scenes but you don't need to bother with the details yourself. The end result is that you do not need to bother with the type conversions or escaping special characters that we just talked about.</p>
            <p>Even with PHP the so called <a href="../../mysql/mysqli.php">improved mysql api (mysqli)</a> let's you work with prepared statements and eliminates the need to escape special characters but the whole idea is to move away from mysql in the first place.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Mysql to postgres migration tool

<p>When you are inserting data into a table that has a generated id you often need to know what the generated id was. With mysql it's real easy, just use the <code>last_insert_id()</code> function. When you enter data into a mysql table that has an <code>auto_increment</code> field you can then make use of this method call retrieve the identifier that was generated. If you happen to be using PHP the job is even easier &nbsp;because you can retrieve this number <code>mysql_insert_id()</code> function of PHP without needing an extra query. At first glance it's counterpart in <a target="_blank" rel="noopener noreferrer nofollow" href="https://www.postgresql.org/">postgres</a> seems to be the <code>currval()</code> function. But appearences can be deceptive. You don't need to pass in any parameters at all to <code>last_insert_id()</code> or the <code>mysql_insert_id()</code> ; &nbsp;the database always returns the id associated with the last executed insert query for that connection. <code>currval</code> on the other hand takes the name of the related <em>sequence</em> as a parameter. So it seems the &nbsp;only way to determine the insert id in pg is by inspecting the database meta data - a cumbersome task at the best of times; it takes quite a bit of effort and querying to get the task done. On the other hand as a programmer you already know the relationship between the sequence and the column so it's a simple matter to create your own function that can return the id for you. In fact it's a mere 5 lines of code.</p><pre><code>CREATE FUNCTION add_user(char,char,integer) RETURNS BIGINT AS '
   INSERT INTO users(userName,userPassword,userStatus) VALUES($1,$2,$3);
   SELECT CURRVAL(''users_userid_seq'');
' LANGUAGE SQL;</code></pre><p> Instead of executing an&nbsp;<em>Insert</em> query as you might have done with mysql you do a <em>SELECT</em> when you want to add data to your table. Sounds counter intuitive? See this example:</p><p><code>SELECT add_user('raditha'.'123',1);</code></p><p> The select is actually using a stored function that does an insert! As with any other <em>select</em> you are presented with a resultset and it contains the value that was last generated by the sequence <em>for the current connection to the data</em> Interestingly when talking to an MS sql server with JDBC you always get back a result set even with an INSERT query if the insert results in the auto increment field been updated. You can do the same by carrying out a select on <em>@@id</em> Follow Up: see the <a target="_blank" rel="noopener noreferrer nofollow" href="../../postgres/index.html">mysql 2 pgsql migration guide</a></p>

last_insert_id or currval

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>If you made the biggest mistake of your life and signed up with <a href="000378.html">AITCOM - the dedicated server failure company</a> you are probably wondering how you can cancel your account. The only way is to cancel your credit card.</p>
            <p>Like many other before me I obtained an account with them without <a href="388.html">googling about them first</a> the server was never accessible and I sent countless emails asking that the acount be cancelled. They sent me the same canned reply over and over again about some wierd procedure to cancel the account.</p>
            <p>According to that I have to login to one of their sites (username / password never given) and then follow up with a phone call (phone never answered). This seems to be they same procedure that others had been asked to follow as well.</p>
            <p>Despite my not using the service and repeated cancellations requests the made <a href="446.html">two unauthorized chargs on my credit card</a>.In desparation I had my card cancelled. Afterwards they had <a href="491.html">attempted to charge it again</a> and since it failed they had finally suspended (not cancelled my account). According to their mail it's all my fault.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

How to cancel your AITCOM account

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Ouch. We have had two server outages in two days. It's a painfull affair for a small company like ours. Our ISP <a href="http://www.webquarry.com/">webquarry</a> was very prompt as always in restarting the server but they are as puzzled as we are as to why this keeps happening.</p>
            <p>The latest crash happened just after we released <a href="http://www.radinks.com/changes.php">version 1.31</a> of our <a href="http://www.radinks.com/sftp/">SFTP Applet</a> - fortunately it happened before we made the announcement so it isn't as bad as it might have been.</p>
            <p>I don't think Rad Inks is big enough (yet) to warrant denial of service attacks but yet for the last three or four months we have had the server die on us during times of peack usage. Interestingly this seems to happen mostly during the first weeks of the month. Could it be that all the spiders are making frequent visits at this time?</p>
            <p>The Spiders of the reputable search engines are all polite blokes. None of them would dream of accessing the sites fast enough to cause the servers to overload but what may be happening is that several bots may be crawling through the sites at the same time then perhaps the load could become a might too much.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Server Down

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>The second pitfal in <a href="486.html">migration from mysql to postgresql is the auto_increment</a> feature for integer types in mysql - the first of course is the <a href="000487.html">password() function</a>.</p>
            <p>In other databases you have to create a sequence and use the nextval() to specify that a field is to be auto incremented. While this may appear to be more cumbersome than the mysql way - it certainly gives greater control. In postgresql there is a shortcut; it allows you to define a field as SERIAL and the sequence is automatically created for you.</p>
            <p>thus <em>messageId int(11) NOT NULL auto_increment</em> in mysql becomes <em>messageId SERIAL</em> in postgres - and that's quite a bit less typing than what you have to do in mysql.</p>
            <p>Another point worth mention about the column definition is that mysql has several types of integers and with any of them you can define the number of bytes to be used. Even though postgress has a <em>smallint</em> and <em>bigint</em> in addition to the standard <em>integer</em> you will rarely need to use anything other than <em>integer</em> and due to a quirk in type casting if you use smallint your query may not make use of the indices as you expect.</p>
            <p>Another datatype that will cause you trouble is the blob - blobs are called <em>bytea</em> in postgresql while in mysql you find them in many flavours including <em>mediumblob</em> and <em>blob</em></p>
            <p>follow up: Aug 05 , 2004 - <a href="000488.html">Retrieving the last insert id</a></p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

mysql to postgres and auto_increment

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p><a href="000378.html">AITCOM is a pathetic excuse for a web hosting company</a>. Last month they <a href="446.html">wrongfully charged</a> me to the tune of $150 - (at the time I thought they had charged me just $69 but turned out to be $150!!). This wasn't <a href="425.html">the first time that they did it either</a></p>
            <p>Apperently they have tried to charge me again but finding the credit card details in their file is no longer valid they have the cheek to send me an invoice by email for a service that I have never used. In my book that makes them as straight as a rolled up network cable. According to the three strikes you are out rule they should be put away for a long time.</p>
            <p>Of course I got into the whole mess by not doing a proper research before I gave them my credit card number. I have failed to come up with any positive comments on AITCOM except those posted by their employees or those who operate hand in glove with them.</p>
            <p>Tales of woe are dime a dozen - If you are planning to go with AIT check out some of these before hand:</p>
            <p><a href="http://www.aitsucks.net">AITCOM (SUCKS) NET</a></p>
            <p><a href="https://www.raditha.com/http://www.webhostingtalk.com/showthread.php?threadid=268534">Web hosting talk thread on AIT</a></p>
            <p>and you will find a few more <a href="388.html">here</a></p>
            <p></p>

        </div>
    </div>


</body>

</html>

AITCOM - Dedicated wrongfull billing service

<p>Many developers particularly those who work in a LAMP environment use the password() function of mysql to encode the passwords that they store in their database. I am not talking even about mysql account password, but passwords as saved by authentication app. The password  function is not supposed to be very secure but people use it anyway - in fact I have also been guilty of this crime at times.</p><p>On the other hand I have seen $1000 applications that store the password in clear text. These poor losers argue that if someone gains unauthorized access to your database you are in big trouble anyway. That is not the point. Passwords should be one way hashed to protect them from friendly eyes as well as those of the intruders. You hear about passwords belonging to thousands of users being released on underground websites all the time. Most people use the same password on multiple sites if compromised once means compromised everywhere.</p><p>No self respecting system administrator or DBA will want to take a peek at your password on purpose but sometimes when doing routine administration it's simply impossible not to see the contents of the password field if it's not encrypted - as they say lead us not into temptation.</p><p>In any event <a target="_blank" rel="noopener noreferrer nofollow" href="https://www.postgresql.org">postgres</a> does not have the same password() function that you find in mysql so you have to encode the password yourself. Now if we had all done things right the first time our PHP code would do the encryption for us instead of relying on the mysql function. Since programmers are lazy by nature it's not done that way.</p><p>Changing your PHP script not to use the password function is is a simple task. You only need one line of code to create an MD5 hash in php - not so with java. Though it is said that you can program with java with little knowledge of what lies under the hood - cryptography is one exception. In this case it is PHP that hides all the gory details while java assumes that you would know how the internal combustion engine worked.</p><p>March 23, 2023 update: MD5 has been considered insecure for a long time (I just didn't get around to updating this blog for many years)</p>

Mysql , postgres and passwords

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>The time has come for the release of version 1.31 of the java <a href="http://www.radinks.com/sftp/">SFTP applet</a> Though I keep referring to the applet there are in fact two different clients. One is the applet that runs in your browser and the other is a standalone client that runs in your desktop.</p>
            <p>What's special about an SFTP applet is that until we came along people used to think this kind of thing was impossible to be done with java - since then a couple of other SFTP applets have appeared on the scene as well as a few FTP applets (we have our own as well.)</p>
            <p>The previous release of Rad SFTP was more than a month ago and I like to think that reflects the growing maturity of the software. In fact the applet will be one year old in August. We plan to celebrate by starting work on version 2.0.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

SFTP

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>In recent years Many a developer has been spoilt by the ease of installing and using mysql. Thus instead of learning SQL they have learnt mySQL - Many of those who have fallen into this pitfall would happen to be PHP developers.</p>
            <p>Fortunately for me when I first learned SQL I learned ANSI SQL and not mysql - even though most of my practising was on <em>my</em>sql database. That was around five years ago. Since then I have done a bit of work with M$ SQL server and worked for a couple (bad) days on an <a href="https://www.oracle.com">Oracle DB</a>. Other than that it has been mysql all the way.</p>
            <p>Now I find myself working with <a href="https://www.postgresql.org/">PostgreSQL</a> and I am having to relearn a few things because of the bad habits aquired by working on mysql. Today I moved one of my databases from mysql to postgres and found myself spending a lot of time RTFMing.</p>
            <p>So what made me switch from being an advocate of mysql for many years to an advocate of postgresql - a database that don't even know all that well? The short answer is mysql licensing issues.</p>
            <p>The mysql AB intepretation of the <a href="https://www.gnu.org">GPL</a> is radically different from everyone else's. GPL is bad enough as it is being the worst of the open source licenses. That's why they should never have switched the client libraries from LGPL to GPL in the first place. If mysql's intepretation of the GPL is valid, Micro$oft's outlook mail client would have to be open source too.</p>
            <p>The reason that mysql is so popular is that developers - PHP developers in particular banded around it, answered messages in newsgroups, put up tutorials and blogged about it. Mysql ab seems to have turned their back on this community and that popularity will wane.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

mysql to Postgres

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I was under the impression that <a href="http://www.cstr.ed.ac.uk/projects/festival/download.html">Festival</a> is abandonware - apparenlty not. They have released a new version labeled as 1.95 (2.0 beta). According to my reconing this is the first major release in around three years.</p>
            <p>I made an immidiate download and started to read the documentation prior to installing it. Unfortunately as often happens with open source projects the documentation has not kept pace with the code. You might have expected festival to be different because it's created by a team from the Edinburg university.</p>
            <p>Professors (usually) seem to care more about the documentation than the code itself. Apperently this is an exception. Good to see professors with common sense for a change.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

A new festival

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>How about that for a function name? I kid you not. It's a function name from the <a href="http://wikipedia.sourceforge.net">mediawiki software version 1.30</a>.</p>
            <p>They have apparently done a lot of rewriting to improve the code and as the function name itself suggests to do away with a lot of globals that are being used in the 1.2x version. They also seem to have a very strong belief in descriptive function names.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

getContentWithoutUsingSoManyDamnGlobals()

<p></p><p>It seems my efforts to <a target="_blank" rel="noopener noreferrer nofollow" href="463.html">configure PHP 5 with java 5 support</a> was a complete and utter waste of time. <a target="_blank" rel="noopener noreferrer nofollow" href="422.html">Java 5 or any version of java</a>, for that matter cannot be used with <a target="_blank" rel="noopener noreferrer nofollow" href="../../php/php5.php">PHP 5</a>. It's perhaps more accurate to say the java extension for PHP 5 is conspicuous by it's absence.</p><p>Since PHP 5 is said to be so much better than PHP 4 in every aspect I didn't bother to find out about the availability of the java extension before my efforts at installation. Last week' attempts to install PHP5 were unsuccesfull due to a <a target="_blank" rel="noopener noreferrer nofollow" href="307.html">conflict with turck mmcache</a>. However at the time I was working with a beta of PHP5. Yesterday I tried with the 'gold release' and managed to have it installed without any problems at all. However the java extension did not show up with a call to <em>phpinfo();</em></p><p>It was then that I examined the ext/ folder to find that the java extension is missing, then I thought it might be available at PECL but I was wrong again. Finally I did what I should have done right at the start; visited <a target="_blank" rel="noopener noreferrer nofollow" href="http://'www.zend.com/">zend.com</a> and did some reading only to find that the java extension has been 'temporarily' dropped from PHP 5.</p><p>So much for the tremoundous improvements.</p><p></p><p></p>

Java 5 and PHP 5 the verdict

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Brown headed barbets are common visitors to my small garden. My latest effort in open source software development is named Barbet in honour of this bird.</p>
            <p>barbet is a database administration tool written in java - or perhaps it would be more accurate to say a database administration tool being written in java. There is less than 1000 lines of code at the moment bu decided to make a release since they say 'release early release often'.</p>
            <p>It's accessibly from <a href="http://www.radinks.net/barbet/">radinks.net</a>. So how come it's not released through this site (<a href="../../index.html">raditha.com</a>)? I get paid for releasing it through radinks.net! what more can you ask for.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Barbet

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Sometime ago I made a <a href="../../php/ftp/index.html">PHP class</a> that was a (partial) implementation of the File Transfer Protocol (FTP) since then I decided to hand it over to <a href="http://www.radinks.net/">Rad Inks</a>.</p>
            <p>This week I made some improvements to it and now we have the start of a <a href="http://www.radinks.net/ftp/">PHP FTP Client</a> the difference of course is that this time around I got paid for working on it.</p>
            <p>Since the mantra of open source development is release early release often, the FTP class itself was released quite sometime ago and has been lying dormant until this week when I blew the dust off and started on the client.</p>
            <p>This PHP ftp client will be released through <a href="https://sourceforge.net/projects/radnetwork">source forge</a> today.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

PHP FTP Class

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I have been doing a bit of reading about semaphores and shared memory for the last few days, in the hope of making an improvement to <a href="../../megaupload">Mega upload</a>. Unfortunately it does not look like a good solution.</p>
            <p>The reason is the String allocation of perl is said to be unpredictable - at least according to the perlipc man page. There are enough problems when interacting between perl and PHP as it is so the last thing I need is another complication.</p>
            <p>I have aslo thought about using named pipes which may be superior to the current file locking mechanism in use. However with named pipes the process that writes to it will block until the second process is ready to read in the data so it will slow down the upload.</p>
            <p>As they say if it's not broken don't fix it.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Shared Memory

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>That heading is misleading. I am not talking about HTTP or FTP error codes. Today I made my 500th post at the <a href="http://forums.devshed.com">Devshed forums</a>. 499 of these postings have been in reply to messages requesting help on varius topics - the one exception was when I posted a question on the AMR codec ( which went unanswered).</p>
            <p>The bulk of the messages were made in the java message board but I made a usefull contribution in the PHP board as well - which if of course in addition to my contribution in the php list where I repond to around 40 questions each month.</p>
            <p>In fact someone who has been lurking in that board invited me to moderate the PHP forums that they operate at : <a href="http://www.weberforums.com/">weberforums.com</a>. I haven't decided wether to accept it or not (probably not I don't have the time). I did visit the site and take a look at the boards on a couple of occaisions after that but there weren't many unaswered messages so I couldn't make myself usefull. I am gonna watch it for a few days more before making the final decision.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

500 Responses

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I came across an java database administration tool for postgresql called <a href="https://www.raditha.com/~raditha/blog/archives/000347.html">XPG</a>. However I immidately ran into problems when attampting to compile it. That's because the source files contains non ASCII characters - since that character sequence is part of the authors name, it's fair enough!</p>
            <p>Apparently it compiles fine under java.4 but fails only with java 5 (1.5.0), at any rate someone had filed a <a href="http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=5046139">bug report</a> as such. The developers of the SDK has claimed it's not a bug but to use the <em>-encoding</em> flag for <em>javac</em>.</p>
            <p>So that's what I did, edited the Makefile for XPG and added <em>-encoding c232</em> as a parameter to <em>javac</em> and the error message (<em>unmappable character for encoding UTF8</em>) disappeared. So if you are having trouble compiling xpg that's something you might want to try.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Java admin tools for Postgres

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>If you are one of those developevers who happen to less than impressed by the feature set of mysql you will be terribly disapointed with <a href="https://www.sqlite.org">sqlite</a>. Then again sqlite isn't really intended to be a replacement for mysql. It's more suited as a lightweight database for dekstop applications or very small sites.</p>
            <p>Perhaps it's not a bad analogy to say Sqlite will become for linux developers what access has been for developers on the windows platform. In other words it's a simple and easy to use data storage system with almost zero installation and administration input requiremed.</p>
            <p>One of the drawbacks of sqlite is that it is weakly typed meaning it does not recognize types (so why is a PHP programmer complaining?). All the data is stored as null terminated strings. That feels kind of like those small storage systems that we did about 10-15 years ago with C.</p>
            <p>Those days if you looked at ten different programs they would have ten different storage systems. Often they used fixed length records but sometimes you ran into null terminated strings. You rarely find programs like that nowadays with access more often than not been the backend (at least on win32 systems).</p>
            <p>Getting back to the null termination issue it effectively rules out the use of blobs in sqlite. Well you can use them if you want to but you have to suitably encode the data first (urlencode would be fine) so that the 'x0' if present get's a different representation. Having said that I have never been a big fan of blobs, most times all you need to do is to save the filename only in the database.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

sqlite

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>So what good are the shared memory functions? well you can use them to share data between different PHP scripts or even with other processes running in your server. Very usefull if you know how to make use it. The trouble is I don't. Shared memory functions are an area that I never looked at before.</p>
            <p>After yesterday's fiasco with <a href="463.html">PHP 5 and Java 5</a> I decided to investigate turck MMCache a bit further and found references to Shared Memory and semaphores, that set me wondering If I might use it in <a href="../../megaupload">mega upload</a> to improve it's performance.</p>
            <p>In any event I had to recompile my php installation to try to accomodate java 5 with php 4 and in the process I decided to add<em> --enable-sysvshm</em> and <em>--enable-sysvsem</em> to the configure script. Some of the user contributed comments in the PHP manual were really usefull in getting started. Just a short while ago I managed to write hello world using shared memory. Here is what it looks like</p>
            <p>shm1.php - saves a variable in shared memory.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

PHP and Shared Memory

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>Does PHP 5 work with <a href="422.html">java 5</a>? I decided to find out. The first step is of course the configuration, the last time <a href="../../php/php5.php">I installed PHP 5</a> I installed both mysql and <a href="../../php/php5.php">mysqli</a> - this time I didn't want to bother with mysql extension so instead of that added --with-java</p>
            <p><code style="margin:15px; font-color:#000033">./configure --enable-xslt --with-xslt-sablot --with-apxs=/usr/local/apache/bin/apxs --with-gd --with-zlib --with-mysqli=/opt/DB/mysql-standard-4.1.1-alpha-pc-linux-i686/bin/mysql_config --with-java</code></p>
            <p>The compilation had the usual <a href="../../linux/krb5.h.php">missing kerberos problem</a> (I am <a href="https://www.raditha.com/blog/archives/000344.html">using apache ssl</a>). After setting the env variables the compilation went smoothly, so did the install now comes the part where I have to edit the php.ini file. I simply copied across the entries from the ini file for my PHP 4 installation and edited the entries so that they now correspond to Java 5 instead of java.4.</p>
            <p>The server startup failed; segfaulted because of a conflict with the <a href="307.html">installed turck MMCache</a> suprisingly no matter what I did (ranging from updating my mmcache version, cleanip up the mmcache objects, trying apache without the SSL etc) the server completely refused to start up. MMCache refuses to allow it self to be removed.</p>
            <p>So now I am going to leave this task in my extremely crowded back burner for the moment and going to take a look at running <a href="https://www.raditha.com/blog/archives/000357.html">PHP 4.3.x with Java 5</a>. In the meantime I have to figure out why turck is playing hard to get and to see if it can ever be made to work with PHP 5.</p>
            <p><a href="000474.html">Follow up: July 24, 2004</a></p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

PHP 5 and Java 5

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I just upgraded from mozilla 1.6 to 1.7 but unfortunately it looks and feels pretty much like what 1.6 used to feel. I don't mind not having a lot of remarkable features to write home about there was one annoying bug in 1.6 which I hope they have done away with in the new version.</p>
            <p>In 1.6 and older versions sometimes when after sending a message the mail client will choke on copying it to the sent folder. It does not matter whether the sent folder is local or on your IMAP store it still chokes. What actually happens is the that 'copying message to sent folder' dialog never goes away. Once I left it running overnight to see what the result would be and in was still running.</p>
            <p>There is a variation as well. Occaisionaly the operation would time out and the dialog box will change to say the copy operation failed. Either way the only way to save your copy would be to ask the mail cilent to save the message as a draft and then move it manually to the sent folder. Less than ideal and didn't always work.</p>
            <p>Hopefully this bug has been eliminated but can't say for sure because it was pretty hard to reproduce even with the older versions. So let's wait and see.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Mozilla 1.7

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>There seems to be quite a lot of activity in the jabber lists of late, both the jadmin and jdev lists and I still haven't got a 2.0 installation! It's been more than an year since I was last involved in a jabber related project and it does look like I will have to do a lot of relearning specially if I want to answer any questions posted on the lists.</p>
            <p>I used to be one of the major contributors to the jadmin list about two years ago and made a noteworthy contribution to the jdev list as well. Since last year though most of my contributions have been in the php list. I often find my self among the top 10 contributors (none of my postings are questions, (though they may not always be the answer the poster wanted to hear)).</p>
            <p>I am just about to notch up my 500th posting at the <a href="http://forums.devshead.com">devshed forums</a> after that I am going to spend a bit of time at least reading the jabber posts to get my hand in.</p>
            <p>Now you may ask why do I (and others like me) spend so much time in posting in message boards. I guess there is a bit of ego involved everyone likes to show off. The main reason though is that no matter how much experience you have, there is always something new to learn and message board postings occaisionaly dig up some real nuggets.</p>
            <p>Another important aspect is that these postings give you a chance to see what kind of technologies other developers are interested in. That matters a lot more than the hyped up press releases distribued by software companies.</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>

Jabber List Reactivated

<html>

<head></head>

<body>
    <div class="item">
        <div class="item-header">



        </div>

        <div class="item-body">

            <p></p>
            <p>I have been having an on again off again affair with postgreSQL for quite some time. My first run in with postgres was in the year 2001. At the time I was just interested in seeing what it had to offer and how it behaved like many others I was under the wrong impression that it's mostly used in academic work.</p>
            <p>If you have been using mySQL you are no doubt as distressed as I am about the license changes and backtracking by mysql ab. Now they are saying mysql client library can be bundled with PHP again but who knows what they may say tommorow? All in all it's time for the love affair that PHP developers have with mysql to end.</p>
            <p>As with thousands of others mysql became my preffered database because it is so easy to install and use. The fact that the mysql documentation is readable certainly helped. The very fact that postgres is a much superior database means that there is a much steeper learning curve involved.</p>
            <p>While mysql can be installed easily even by developers with limited knowledge of system administration the same cannot be said about postgres. Similarly the first few steps, creating databases, adding tables seems a wee bit more difficult with postgres when compared to mysql.</p>
            <p>My gripe today though is about retrieving meta data using the JDBC driver. If you are using the psql command line client you can retrieve information about a table with the <em>d tablename</em> command. Unfortunately that command cannot be used from the JDBC driver and <em><b>desc</b>ribe tablename</em> syntax is not supported in either.</p>
            <p>Curiosly the pgsql api in PHP has a function that can retrieve meta data for you. So how come it's supported in one but not in the other?</p>
            <p></p>
            <p></p>

        </div>
    </div>


</body>

</html>