Re: (Never?) Kill Postmaster? - Mailing list pgsql-general

From Tom Lane
Subject Re: (Never?) Kill Postmaster?
Date
Msg-id 17425.1194971329@sss.pgh.pa.us
Whole thread Raw
In response to Re: (Never?) Kill Postmaster?  (Christian Schröder <cs@deriva.de>)
List pgsql-general
=?ISO-8859-1?Q?Christian_Schr=F6der?= <cs@deriva.de> writes:
> What remains inscrutable to me is why this problem did not arise
> earlier. I cannot remember any changes that I have made to the system
> recently. Or maybe it has been a defective update of the glibc? I did
> not see in the bug report any hint about when this bug has been
> introduced.

It's quite recent actually --- according to
http://sourceware.org/cgi-bin/cvsweb.cgi/libc/intl/dcigettext.c?cvsroot=glibc
the problem was introduced into upstream sources on 24-Sep.  I was a bit
confused yesterday because I'd been looking at another lock used by the
same routine that's been there much longer; but that one is initialized
in a safe way so it doesn't have the problem.  It's the new usage of the
conversions_lock that puts _nl_find_msg at risk.  So even though the
underlying problem has existed for years, it didn't manifest, at least
not right there.

So I suppose that SuSE is tracking upstream closely and you recently
updated glibc ...

> P.S.: Can I change the root password again?

Oh, yeah, I'm done.  Thanks for letting me poke at it.

            regards, tom lane

pgsql-general by date:

Previous
From: Scott Ribe
Date:
Subject: Re: Linux v.s. Mac OS-X Performance
Next
From: Decibel!
Date:
Subject: Re: Using generate_series to create a unique ID in a query?