ECPG regression failures on OpenBSD

Started by Stefan Kaltenbrunnerover 19 years ago10 messages
#1Stefan Kaltenbrunner
stefan@kaltenbrunner.cc

After upgrading some of my buildfarm members to the latest version of
the buildfarm script both OpenBSD boxes startet to fail the ECPG
regression tests:

http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=emu&dt=2006-09-05%2002:35:02

and

http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=spoonbill&dt=2006-09-04%2023:50:05

Stefan

#2Michael Meskes
meskes@postgresql.org
In reply to: Stefan Kaltenbrunner (#1)
Re: ECPG regression failures on OpenBSD

On Tue, Sep 05, 2006 at 08:00:12AM +0200, Stefan Kaltenbrunner wrote:

After upgrading some of my buildfarm members to the latest version of
the buildfarm script both OpenBSD boxes startet to fail the ECPG
regression tests:
...

complex/test2 gets a segmentation fault on both machines. Could you try
running it under gdb to see where it segfaults? I will try to get a hand
on an OpenBSD machine myself, too

Michael
--
Michael Meskes
Email: Michael at Fam-Meskes dot De, Michael at Meskes dot (De|Com|Net|Org)
ICQ: 179140304, AIM/Yahoo: michaelmeskes, Jabber: meskes@jabber.org
Go SF 49ers! Go Rhein Fire! Use Debian GNU/Linux! Use PostgreSQL!

#3Stefan Kaltenbrunner
stefan@kaltenbrunner.cc
In reply to: Michael Meskes (#2)
Re: ECPG regression failures on OpenBSD

Michael Meskes wrote:

On Tue, Sep 05, 2006 at 08:00:12AM +0200, Stefan Kaltenbrunner wrote:

After upgrading some of my buildfarm members to the latest version of
the buildfarm script both OpenBSD boxes startet to fail the ECPG
regression tests:
...

complex/test2 gets a segmentation fault on both machines. Could you try
running it under gdb to see where it segfaults? I will try to get a hand
on an OpenBSD machine myself, too

will try to get a backtrace soon - if you are trying to do your own
testing keep in mind that both boxes of mine do run with special
malloc-settings (as in FGJZ) as discussed in:

http://archives.postgresql.org/pgsql-hackers/2005-06/msg00817.php

while I'm not sure yet that those are causing the errors to show up it
seems quite likely since they tend to catch hidden memory allocation errors.

Stefan

#4Chris Mair
chrisnospam@1006.org
In reply to: Stefan Kaltenbrunner (#3)
Re: ECPG regression failures on OpenBSD

After upgrading some of my buildfarm members to the latest version of
the buildfarm script both OpenBSD boxes startet to fail the ECPG
regression tests:
...

complex/test2 gets a segmentation fault on both machines. Could you try
running it under gdb to see where it segfaults? I will try to get a hand
on an OpenBSD machine myself, too

will try to get a backtrace soon - if you are trying to do your own
testing keep in mind that both boxes of mine do run with special
malloc-settings (as in FGJZ) as discussed in:

http://archives.postgresql.org/pgsql-hackers/2005-06/msg00817.php

while I'm not sure yet that those are causing the errors to show up it
seems quite likely since they tend to catch hidden memory allocation errors.

Hi,

I've just upgraded guppy, the only other OpenBSD machine that builds
head to the latest buildfarm version.

I don't have any special malloc settings. When guppy finishes, we will
see what happens.

Bye, Chris.

--

Chris Mair
http://www.1006.org

#5Michael Meskes
meskes@postgresql.org
In reply to: Stefan Kaltenbrunner (#1)
Re: ECPG regression failures on OpenBSD

On Tue, Sep 05, 2006 at 11:58:41AM +0200, Stefan Kaltenbrunner wrote:

#0 0x0add8c83 in replace_variables (text=0x7e709780 "select name, born,
age, married, children from meskes where name = ? ") at prepare.c:42
ptr = 0x7e70a000 <Address 0x7e70a000 out of bounds>
string = 0 '\0'

I think I found this one. Will commit as soon as I finished my tests and
the parser update.

Michael
--
Michael Meskes
Email: Michael at Fam-Meskes dot De, Michael at Meskes dot (De|Com|Net|Org)
ICQ: 179140304, AIM/Yahoo: michaelmeskes, Jabber: meskes@jabber.org
Go SF 49ers! Go Rhein Fire! Use Debian GNU/Linux! Use PostgreSQL!

#6Michael Meskes
meskes@postgresql.org
In reply to: Stefan Kaltenbrunner (#1)
Re: ECPG regression failures on OpenBSD

On Tue, Sep 05, 2006 at 08:00:12AM +0200, Stefan Kaltenbrunner wrote:

http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=emu&amp;dt=2006-09-05%2002:35:02

and

http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=spoonbill&amp;dt=2006-09-04%2023:50:05

Hmm, for some reason the x86 machine doesn't give a
PGTYPES_NUM_UNDERFLOW error when it should. The Sparc correctly throws
that error as do Linux based x86 machines.

Michael
--
Michael Meskes
Email: Michael at Fam-Meskes dot De, Michael at Meskes dot (De|Com|Net|Org)
ICQ: 179140304, AIM/Yahoo: michaelmeskes, Jabber: meskes@jabber.org
Go SF 49ers! Go Rhein Fire! Use Debian GNU/Linux! Use PostgreSQL!

#7Michael Meskes
meskes@postgresql.org
In reply to: Stefan Kaltenbrunner (#1)
Re: ECPG regression failures on OpenBSD

On Tue, Sep 05, 2006 at 01:08:48PM +0200, Stefan Kaltenbrunner wrote:

FYI: complex/test2 fails on lionfish (my linux/mipsel box) too:

http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=lionfish&amp;dt=2006-09-05%2005:30:07

This needs some debugging. The output difference comes from an indicator
not correctly set/evaluated. To get more infos I need to get access to
such a machine.

Michael
--
Michael Meskes
Email: Michael at Fam-Meskes dot De, Michael at Meskes dot (De|Com|Net|Org)
ICQ: 179140304, AIM/Yahoo: michaelmeskes, Jabber: meskes@jabber.org
Go SF 49ers! Go Rhein Fire! Use Debian GNU/Linux! Use PostgreSQL!

#8Chris Mair
chrisnospam@1006.org
In reply to: Chris Mair (#4)
Re: ECPG regression failures on OpenBSD

will try to get a backtrace soon - if you are trying to do your own
testing keep in mind that both boxes of mine do run with special
malloc-settings (as in FGJZ) as discussed in:

http://archives.postgresql.org/pgsql-hackers/2005-06/msg00817.php

while I'm not sure yet that those are causing the errors to show up it
seems quite likely since they tend to catch hidden memory allocation errors.

Hi,

I've just upgraded guppy, the only other OpenBSD machine that builds
head to the latest buildfarm version.

I don't have any special malloc settings. When guppy finishes, we will
see what happens.

Hi,

after a spurious run due to a config mistake after upgrading (sorry),
guppy just now completed a new run and also failed on "make ecpg check":

http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=guppy&amp;dt=2006-09-05%
2010:02:05

Michael, if you want shell access to guppy, just contact me privately.
Warning: guppy too, is somewhat dated (1:10 hours for the make step) :/

Bye, Chris.

--

Chris Mair
http://www.1006.org

#9Stefan Kaltenbrunner
stefan@kaltenbrunner.cc
In reply to: Michael Meskes (#7)
Re: ECPG regression failures on OpenBSD

Michael Meskes wrote:

On Tue, Sep 05, 2006 at 01:08:48PM +0200, Stefan Kaltenbrunner wrote:

FYI: complex/test2 fails on lionfish (my linux/mipsel box) too:

http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=lionfish&amp;dt=2006-09-05%2005:30:07

This needs some debugging. The output difference comes from an indicator
not correctly set/evaluated. To get more infos I need to get access to
such a machine.

hmm I might be able to arrange access to lionfish - but that box is slow
(it needs between 5 and 6 hours to finish a buildfarm run).

Stefan

#10Noname
chrisnospam@1006.org
In reply to: Chris Mair (#8)
Re: ECPG regression failures on OpenBSD

Michael, if you want shell access to guppy, just contact me privately.
Warning: guppy too, is somewhat dated (1:10 hours for the make step) :/

Michael,
did you receive my private mail yesterday?
(just want to make sure it wasn't blocked by an overzealous spam filter)

Bye, Chris.

--

Chris Mair
http://www.1006.org