31-Dec-2008

PHP issues
I found that the new PHP engine seems to be somewhat case sensitive….
When trying to add new content to the tracks blog, I entered the location as “Tracks” and that failed – causes the location not found. But stating “tracks” – all lowercase – works. At least, at that time it meant a difference. Once things have been accessed, case doen’t seem to matter anymore: I cannot reproduce it at will now.
The same problems happened with this blog. Once in a while, nothing seems wrong and all works fine, at other times I get “stack overflow” at some places, and MySQL may (!) fail accessing RSS data; This is found in the webserver log:
WordPress database error MySQL server has gone away for query UPDATE sm_options SET option_value = 'O:9:\"MagpieRSS\":19:{s:6:\"parser\";i:0;s:12:\"current_item\";a:0:{}s:5:\"items\
It happens on any of the admin screens, I guess this is a MySQL problem…
The same may happen using SET in stead of UPDATE in the SQL statement. It causes errors on accessing the server, but refresh could well return the page.

So it’s not as stable I would expect. Big problem: Re-starting the older version failes with ACCVIO, no matter what I try….But normal use seems to work pretty well – and fast, once the environment has been setup – so I keep it this way – for now (there is no real alternative 🙁

Others are testing as well, it has been found that UTF-8 is now built-in in the PCRE module. That’s true, since at times I do not get this error in the sysblog admin pages. But other required modules are not included, and these could cause the problems I found including media.

Some digging to do, still, before sending a report to HP.

26-Dec-2008

Windows woes
No answer on the Outlook problems….
I got extra memory and an extra disk for Aphrodite, to be able to install VMWare ESXi. The new disk was prepared to hold this software, and it took some while that I found out I had to use Demeter – the laptop – to access the machine, but what was worse, I couldn’t use the drives on Aphrodite: I could install the system software (Windows XP and Ubuntu Desktop) only using Demeter’s DVD drive. And it took far too long, so finally I gave up, and re-initiated the drive to hold both data partitions I had set up before; the two on the primary disk now became free to do a ree-install of Windows XP and do a Ubuntu intstall. Both succeeded, and have been updated to latest standards. I didn’t install Office – yet. I think I’ll move to Linux for most of the normal office job and keep the Windows installation for whwn I really need it. Or run it in emulation on Linux 😉
Anyway, the alternative installation of Windows now contains all latest free MS stuff: IE8 (beta 2), Silverlight, Live mail and messenger, and of course, McAfee virus protection and firewall. More to come there, since the normal installation has more weird and unexpected issues every time: lost USB devices, problems in connection to other Diana; hangs of Word and Excel. But none of the scanning software: Ad-Aware, Spybot S&D nor Windows Defender find somnething that shouldn’t be there…

25-Dec-2008

Further testing
It seems to be a problem to get the previous (PHP 4.1) version back online; ANY access done fails with an “Access Violation”. Still to find out what causes it, but the environment isn’t too clear anymore so I have to do some cleanup and redesign first. But that didn’t stop me digging soemwhat deeper.

On PHPMyAdmin, I found some more things, it may have to do with some logging. After having stripped some data, this is what is returned, accoring WATCH:


|19:50:20.34 DCL      4324 0001 DCL        READ SYS$OUTPUT %X00000001 420 bytes|

savepath is /tmp<LF>savepath is /tmp<LF>
mode is 384<LF>
datafd is 4<LF>
buf is /tmp/sess_TY4hZWH2RM84kxpNyUOGtHuOO0b<LF>
Status: 0<CR><LF>
X-Powered-By: PHP/5.2.6<CR><LF>
Set-Cookie: phpMyAdmin=TY4hZWH2RM84kxpNyUOGtHuOO0b; path=/phpmyadmin/<CR><LF>
Expires: Thu, 19 Nov 1981 08:52:00 GMT<CR><LF>
Cache-Control: no-store, no-cache, must-revalidate, post-check=0, pre-check=0<CR><LF>
Pragma: no-cache<CR><LF>
Content-Type: text/html; charset=utf-8<CR><LF>
Script-control: X-stream-mode<CR><LF><CR><LF>

|19:50:20.34 CGI      2030 0001 CGI        NOT a strict CGI response!|

The lines above “Status” are the culprits for the final result. I don’t know where these lines come from: PHPSHR, somewhere, or from PHPWASD? These lines are after the system states:

|19:50:18.33 DCL      4553 0001 DCL        CALLOUT begin|
|19:50:18.33 DCL      4324 0001 DCL        READ SYS$OUTPUT %X00000001 16 bytes|
21434749 504C5553 3A207374 72756374                                     !CGIPLUS: struct
|19:50:18.33 DCL      4575 0001 DCL        CALLOUT 16 bytes|
21434749 504C5553 3A207374 72756374                                     !CGIPLUS: struct
|19:50:18.33 DCL      4324 0001 DCL        READ SYS$OUTPUT %X00000001 28 bytes|
24442D43 41303233 31344631 31393137 46343446 34334435 4437352D          $D-CA02314F11917F44F43D5D75-
|19:50:18.33 DCL      4592 0001 DCL        CALLOUT end|

It’s another issue why the status is 0, and not 200, or another.
This is why I wanted to have PHP4 working again, to cross-check….But as said, I’ll have to clear up things a bit.
But this is to be mentioned to Mark, and HP, probably.

I also retried E107 – a framework once installed, as the gererator for the home page, and others. It failed in the provious PHP version, and I cleared the database so I had to redo the installtion. But that fails – due to a SQL problem, still to investigate.

This blog has it’s quirks as well. For the blog itself, as a reader, I don’t think there is much of a problem. Logging in, however, seems to be. At least, the browser tells me it cannot show the requested (admin) page, refresh won’t work either, but on restart, login proved to have been successful, and you can accees the admin pages. That’s a cookie thing, I guess, it can also be a MySQL problem, since the request is checked against the database, and this may give an invalid result.
Need more diggeing here as well.

Outlook problem on Aphrodite
On Aphrodite, I use Microsoft Office 2000 for most of the wordprocessing and spreadsheet work – ans Ootlook to handle e-mail and calendar. I’ve done so for at least 8 years now, until last Monday. Sunday night, I could get my mail from all the servers, but on Monday, Outlook stops loading after displaying the specls-screen – reading millions of bytes, consuming pagefile space – but without getting any further than that. I checked on the Microsoft Office forum, the only reply so far has been checking the .PST files, for an error in one of them may cause a problem for Outlook to start. There was one problem, but repair didn’t change a bit. Repairing the installation, or re-installing it (after having it removed) didn’t either. Finally, I copied the .PST files to Demeter, had to convert them to Office2003 format, and got my mail that way.
Copying these files had to be done using the CIFS share since direct access between the machines seems to be impossible at the moment. It must be something in the CIFS 1.1 configuration, but I didn’t dig into this at the moment.

To bypass the problem, I installed OpenOffice 3.0 as well, but it does not contain the component I was looking for: something that works like Outlook and is able to read (and update) Outlook’s Personal Folders and Address Books (.PST and .PAB files). Well, keep searching….

24-Dec-2008

Patches installed.
Today, I installed a number of patches – after having made a proper image backup of the system disk.
That was no problem, as could be expected.
First tests of PHP
A few weeks ago, I learned a new version of PHP was coming soon. It’s abit delayed,but I was able to put it to thye test. Well, some tests.
To keep the previous version at hand, I renamed the previous directory trees to PHP4 and the newly created one PHP5. From there all updates have been done: PHPWASD that had to be changed a bit.
After reboot, PHP5 was enabled but there were some slight problems in the startup script, that caused PHP_ROOT not be defined. A small change solved that problem. I also forgot to move PHPWASD to the right spot, but once these issues were settled, I could start running tests.
MyPHPAdmin fails. From the start, it gives a non-CGI-conformance reply. Obvious, it doesn’t return the status as a first issue. Weird, since that should haven been returned. Or it’s caused by the MySQL 4.1 engine that came with the PHP package. That’s something to be reconsidered!
WP – the test blog – runs but is VERY slow. Login may fail, but will in the end succeed. Or may return an eror, but WATCH shows the full HTML output, which is valid. And some data is added, which may cause a problem.
This blog had a problem on login – but refreshing the screen would come back with the admin page in the end. Not sure yet what caused it. But where I previously found a message that PCRE lacked UTF-8 support, I get a more severe error:

%SYSTEM-F-STKOVF, stack overflow, PC=00000000002399E0, PS=0000001B
%TRACE-F-TRACEBACK, symbolic stack dump follows

 image     module      routine        line    rel PC           abs PC

 PHPSHR                                    0  00000000001E79E0 00000000002399E0
 PHPSHR                                    0  00000000001E931C 000000000023B31C
 PHPSHR                                    0  0000000000266634 00000000002B8634
 PHPSHR                                    0  00000000001A1AC4 00000000001F3AC4
 PHPSHR                                    0  0000000000189D94 00000000001DBD94
 PHPWASD   PHPWASD     ProcessRequest  44025  0000000000001C58 0000000000031C58
 PHPWASD   PHPWASD     main            43639  00000000000005A8 00000000000305A8
 PHPWASD   PHPWASD     __main          43565  000000000000006C 000000000003006C
                                           0  FFFFFFFF80379CE4 FFFFFFFF80379CE4
%TRACE-I-END, end of TRACE stack dump

and later on, where plugins list is shown, the same error is found.

Second, it’s VERY slow. And where I previously could handle fo9rms nicely within a 12Mb linit, I get Out-of-core messages now and than – on logout, for instance.
Testing getting an image wasn’t changed. After uploading the image, stating the image being included in the post, dows not return to the editor: IE states “The object doesn’t support this property or method”.
But the script is there if you view the source:

<script type="text/javascript">
/* <![CDATA[ */
var win = window.dialogArguments &vbar;&vbar; opener &vbar;&vbar; parent &vbar;&vbar; top;
win.send_to_editor('<a href=\"https://homedesk.grootersnet.nl/wp265/wp-content/uploads/2008/12/3c75_nrao_big.jpg\"><img src=\"https://homedesk.grootersnet.nl/wp265/wp-content/uploads/2008/12/3c75_nrao_big.jpg\" alt=\"Some astronomical image\" title=\"3c75_nrao_big\" width=\"500\" height=\"500\" class=\"size-full wp-image-20\" /></a>');
/* ]]> */
</script>

Well, I found that adding the data between the “” tags would do the trick. With, without these tags.
In the course of the next day, I’ll do some more testing to find out what’s going on.
UPDATE
Another issue just found: Where previously lines in withing <code> tags would wrap if they exceeded the width of the section, it’s not chopped on the right.
But I also found that once the site environment is set up, access to pages is faster than before. But that’s more like the result of Mark’s coding than the PHPSHR module 🙂

10-Dec-2008

Mondesi
Mark Daniel – the author of the WASD webserver and the SoyMail web-based VMS-MAIL agent – has released a new tool for system monitoring, named Mondesi. It offers you an almost real-time view of the load on different resources in the system.
I installed it under the Homedesk page but had trouble getting it to work. But today I found the reason: The mapping of the directories used by the image was never reached: the BASIC.CONF file – included in the homedesk configuration file – contains a mapping of “/*/-/*”, and “/mondesi/-/*” matches. Of course, the files searched for will never be found there. So I placed this mapping in front of the inclusion of basic.conf. And now it works fine:
mondesi screenshot.
Not satisfied? Check it live.
This package can also be used if you use SWS (Apache) or OSU, how to set things up is described. I didn’t try these, though.
Tracks added
It took some time to get it finished, but the first two walks taken around Nijmegen have been published. Partly, at least: just the images, because I didn’t have the trackdata at hand. But to get it that far, I needed to install a new version of SnagIt and Jalbum on the laptop…Both free (that is: I have a license for SnagIt, which is shareware, but Jalbum is free. Completely).
New PHP imminent?
I still need to check out the WP trouble with PCRE without UTF. But is seems a new PHP version is imminent, accoring the WASD mailing list. Wait until that one is out – must be within a few weeks (around Christmas?). That would fit the remark in last year’s bootcamp.