topic Re: PHP abort in Operating System - OpenVMS

PHP abort

Wim Van den Wyngaert — Wed, 19 Jan 2005 05:47:59 GMT

Hi,

I'm trying drupal on php. But when accessing the admin menu I get an abort. See enclosure.

Does anyone knows the functionality of DBGTBKMSG_V72X or what could be wrong in general ?

Wim

Re: PHP abort

Karl Rohwedder — Wed, 19 Jan 2005 06:18:32 GMT

Perhaps you should try a newer kit (ADB074).

Re: PHP abort

Wim Van den Wyngaert — Wed, 19 Jan 2005 06:23:16 GMT

Karl,

Of what ? What is it and which kit ? I'm using VMS 7.3 and I can't relink php ?

Wim

Re: PHP abort

Kris Clippeleyr — Wed, 19 Jan 2005 06:51:57 GMT

Wim,

The DBGTBKMSG.EXE is the TRACE message file.
It seems that your version of PHP is linked with an old one (V72 might indicate the file that came with VMS 7.2 ???).
Maybe you could replace the faulty one with the real one (just copy the SYS$MESSAGE:DBGTBKMSG.EXE to the location where PHP thinks the DBGTBKMSG_V72X.EXE si located, and name that copy thus). Normally DBGTBKMSG is also installed, take care.

Kris (aka Qkcl)

Re: PHP abort

Wim Van den Wyngaert — Wed, 19 Jan 2005 07:16:33 GMT

Kris,

I defined DBGTBKMSG as pointing to the real system file. It accessed the file (set watch) but still it aborted. I used the file dated 17-mar-2001. The v72 dated 4-mar-1999.

But why is it aborting when accessing a simple message file ?

Wim

Re: PHP abort

Kris Clippeleyr — Wed, 19 Jan 2005 07:48:38 GMT

Wim,

But why is it aborting when accessing a simple message file ?

Obvious (stupid) answer:
A routine in PHP_SHR tries to read (reason mask=00) from virtual address 4. And that is AFAIK not permitted. Could that be a programming error?
I know that this doesn't help, but I'm not familiar with (the coding of) PHP.

Regards,

Kris (aka Qkcl)

Re: PHP abort

Wim Van den Wyngaert — Wed, 19 Jan 2005 11:43:44 GMT

All,

I defined the logical pointing to a non-existing file. Result : also access violation. So I guess it must be something after the access to the message file.

But what ?

Wim

Re: PHP abort

Jan van den Ende — Wed, 19 Jan 2005 16:09:28 GMT

Wim,

I have to agree with Kris. A coding error.
The only alternative that I know of (and hard to do at that) is a changed sharable image, in which relevant addresses have shifted.
But on _PROPER_ version behavior, link time info should result in Image Activation "Version Mismatch".
However, on ported software I would not dare to fully rely on that...

In other words: how SURE are you that you have the matching version of the shareble (in this case probably, the message-) image file?

Of course, it also could be a REAL programming error in an address-pointer-manipulating routine. But then, a routine that is only rarely executed, or the error would have been found much earlier.

Sorry to be of no more help...

(and I am now on Grimbergen Triple. Recommandable!)

Proost.

Have one on me.

jpe

Re: PHP abort

Wim Van den Wyngaert — Thu, 20 Jan 2005 01:55:32 GMT

Jan,

Check my last post : without the image I also have the problem.

Keep on running on Tripple and Duvel.

The problem is that I don't find a way of letting PHP tell me what he was exactly executing (a trace of statements).

Wim
Junior-php-er

Re: PHP abort

Wim Van den Wyngaert — Thu, 20 Jan 2005 01:56:11 GMT

Btw : I also tried it with an 7.2 DBG file. Idem.

Re: PHP abort

Bojan Nemec — Thu, 20 Jan 2005 02:28:52 GMT

Wim,

I agree with Kris and Jan that this is probably a coding error (probably the program is passing a null pointer instead of a descriptor).

Abbout the DBGTBKMSG_V72X. This file is accesed after the error. You can try with a small program which causes an acces violation:

.psect code exe,nowrt
.entry accvio ^m<>
movl #0 , r0
movl #0 , (r0)
ret
.end accvio

On my system this access the DBGTBKMSG.EXE. The DBGTBKMSG_V72X seems be part of the FORTRAN distribution (try with Google).

Bojan

Re: PHP abort

Wim Van den Wyngaert — Wed, 30 Nov 2005 09:20:34 GMT

I upgraded to PHP 1.3.
The abort remains present.

Wim

Re: PHP abort

Mark Daniel_2 — Wed, 30 Nov 2005 10:28:55 GMT

This 'looks' similar to

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=978328

which has been suggested (after some analysis) by the support centre (John Gillings) to be reminiscent of an uninitialised variable. An ACCVIO of course suggests a NULL pointer being dereferenced.

The above thread is ongoing and the next step is (for me :-) to rebuild the CSWS PHP source with traceback so we can try to get some better handle on the PHP source function where it's happening.