Обсуждение: solution for psql segmentation fault ??

greetings postgres gurus:

i've scoured the mail list logs and found many requests for assistance
with this problem:

``i just built the new pgsql release and now psql quits with a
segmentation fault every time i run it''

but can not find any answer - maybe all these people worked it out on
their own or abandoned postgres. does anyone have a hint? 

i'm guessing its something to do with versions of readline/glibc but don't
have a clue what to try upgrading/downgrading or patching to make
everything work in harmony. everything builds without complaint and the
postmaster appears to be running just fine - psql just segfaults.

does anyone have a solution to build psql that worked for them? 

i'm building postgres ~6.5 from cvs on intel/pentium redhat linux boxes
with glibc 2.x (libc6) readline 2.2.1 and linux kernel 2.2.10+ using pgcc
(egcs 1.1.2).

profuse thanks to anyone who may have a moment to suggest some help and
Cc: directly to my email address.

can't wait to try out the 6.5.x postgres features!

--
mailto:sean@savvysearch.com                sean dreilinger, mlishttp://www.savvysearch.com
http://durak.org/sean

sean dreilinger <sean@savvysearch.com> writes:
> i've scoured the mail list logs and found many requests for assistance
> with this problem:
> ``i just built the new pgsql release and now psql quits with a
> segmentation fault every time i run it''

You have?  This is the first I've heard of it ...

Can you provide a gdb backtrace from the corefile?

> i'm guessing its something to do with versions of readline/glibc but don't

If you suspect readline, there's a switch you can give to prevent psql
from using readline --- does psql -n behave any differently?
        regards, tom lane

> i've scoured the mail list logs and found many requests for assistance
> with this problem:
> ``i just built the new pgsql release and now psql quits with a
> segmentation fault every time i run it''

Methinks you are overstating the case a bit; you didn't need to, since
we would have tried to help anyway ;)

> but can not find any answer - maybe all these people worked it out on
> their own or abandoned postgres.

??

> i'm building postgres ~6.5 from cvs on intel/pentium redhat linux 
> boxes with glibc 2.x (libc6) readline 2.2.1 and linux kernel 2.2.10+ 
> using pgcc (egcs 1.1.2).

A RH-5.2/glibc-2.0.x/linux-2.0.36 machine is the "standard" platform
for Postgres. I haven't bumped to the 2.2 kernel, but afaik there are
no fundamental problems. Oleg in Russia was testing bleeding edge
kernels and Postgres some time ago, but I'm pretty sure that things
settled down for him.

The almost deafening silence to your inquiry might indicate that folks
can't see what the problem might be, since they have been successful
themselves. I'm not sure what vintage cvs tree you are using; if it
predates the official v6.5 release then all bets are off since some
newly reported bugs were being fixed (and re-fixed) almost up to the
release date.
                    - Thomas

-- 
Thomas Lockhart                lockhart@alumni.caltech.edu
South Pasadena, California

Tom Lane wrote:
> sean dreilinger <sean@savvysearch.com> writes:
> > i've scoured the mail list logs and found many requests for assistance
> > with this problem:
> > ``i just built the new pgsql release and now psql quits with a
> > segmentation fault every time i run it''

> You have?  This is the first I've heard of it ...

its in the archive, mentioned around the time of each new release, maybe
since 6.2?

> > i'm guessing its something to do with versions of readline/glibc but don't
> If you suspect readline, there's a switch you can give to prevent psql
> from using readline --- does psql -n behave any differently?

readline circa redhat linux 5.1 was installed here, i built
readline-2.2.1-5.src.rpm from updates.redhat.com and psql from the
postgres 6.5 cvs repository worked immediately without recompiling.

this evening i rebuilt the entire postgres distribution, everything runs
great!

> Can you provide a gdb backtrace from the corefile?

if this would be helpful in developing postgres i'm willing to put back
the old readline and crash psql a few more times :-). 

thanks very much to the folks who took time to rsvp with $.02!

--sean


--
mailto:sean@savvysearch.com                sean dreilinger, mlishttp://www.savvysearch.com
http://durak.org/sean