Re: Segfault discovered, possibly due to DBI 1.45: FAIL for Perl 5.8.4/Linux/gcc/DBI 1.45/SQLite 1.07

From: Tim Bunce (Tim.Bunce_at_pobox.com)
Date: 11/04/04


Date: Thu, 4 Nov 2004 09:51:17 +0000
To: Max Maischein <corion@corion.net>, daoswald@adelphia.net


[CC'd to dbi-users as an update and to help others know how best to help.]

On Wed, Nov 03, 2004 at 06:49:10PM +0100, Max Maischein wrote:
> Here's the second Linux run (still crashing), with DBI and DBD::SQLite
> upgraded to 1.45 and 1.07 respectively. I hope the coredumps are what
> you need

No, sorry, the core files themselves are no use to me (without
access to your system). What I really need is a "stack trace" that
you can get from the core file while it's on your system.

The Devel::CoreStack module may help you get a stack trace, but it
ought to simple to get manually. Try running this command:

  gdb perl perl.core # where "perl.core" is whatever your core file is called

then give gdb the "bt" command and tell me what it says.

It would help if both the DBI and the driver are compiled with
the -g option. For the DBI you can do that using "perl Makefile.PL -g".

The driver might not provide a simple way to do that. If not, run
"perl Makefile.PL" then edit the generated makefile and add "-g" to
the OTHERLDFLAGS and OPTIMIZE. If OPTIMIZE has a -O (or similar) flag
then remove it.

The reports from windows indicate that the segfault is happening
in the driver, so compiling the driver with -g before retesting and
getting the stack trace would help a lot.

> ... My Perl 5.6.1 (with DBI 1.45 and 1.07) exhibits the same
> behaviour - I can send those coredumps and Perl -V too if that helps.

It's useful to know that it can be reproduced in 1.42 (March 2004),
but for finding the cause please use 1.45 to get the stack trace.

Tim.



Relevant Pages

  • Re: ZwClose in ndis minport driver crashing
    ... On Windows Vista, enabling the Driver Verifier Miscellaneous Checks will enable Handle Tracing for the System process. ... The stack trace history gets lost if the first Close happened a while back though - only the very recent traces are available. ...
    (microsoft.public.development.device.drivers)
  • Re: BugCheck 0xA when verifier enabled
    ... IoFreeIrp and my problem is resolved. ... or frees the Irp using ExFreePool function instead of IoFreeIrp. ... the stack trace does not point to my driver at all. ... Here's my stack trace on the fault. ...
    (microsoft.public.development.device.drivers)
  • Re: WLK: Rebalance test causes BSOD for my parent
    ... BSOD here too. ... But I got a stack trace as well and it seemed that it ... USB Driver bugcheck, first parameter is USB bugcheck code. ...
    (microsoft.public.development.device.drivers)
  • Re: BugCheck 0xA when verifier enabled
    ... Here's my!analyze dump. ... FOLLOWUP_NAME: MachineOwner ... the stack trace does not point to my driver at all. ... Here's my stack trace on the fault. ...
    (microsoft.public.development.device.drivers)
  • Re: Oracle driver killing jvm
    ... twice a day now with NO Exceptions, no segmentation faults and it appears no core files. ... Anyway we're testing the newest thin driver as I type this but I was wondering if anyone had a way to tell if it was for sure the oracle driver killing the jvm. ... The way to find the culprit would be to start your JVM under a C-level debugger like dbx or whatever one your OS has. ...
    (comp.lang.java.databases)