Mailing-List: contact apreq-dev-help@httpd.apache.org; run by ezmlm
Precedence: bulk
To: Arthur Bergman <sky@nanisky.com>, apreq-dev@httpd.apache.org
Subject: Re: Perl segfault in 5.8
References: <CE2AEDCF-7FA4-11D7-BE1E-000393CB5BC4@nanisky.com>
	<3EB78380.5040504@stason.org>
From: Joe Schaefer <joe+apache@sunstarsys.com>
Date: 06 May 2003 11:11:06 -0400
In-Reply-To: <3EB78380.5040504@stason.org>
Message-ID: <m3u1c82jqt.fsf@sol.sunstarsys.com>
Lines: 27
User-Agent: Gnus/5.09 (Gnus v5.9.0) Emacs/21.2
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii

Stas Bekman <stas@stason.org> writes:

> [redirecting to the right mailing list]
> 
> Arthur Bergman wrote:
> > Hi,
> > We are having a problem with weird segfaults in IO_flush.
> >  Reading old  emails I see this
> > http://mathforum.org/epigone/modperl/zhoolamgheld/
> > Pine.LNX.4.33.0206131508310.955-100000@mako.covalent.net
> > but this does not seem to be applied to libapreq, any reason why?
> > Arthur

ISTR a pair of discussions, this being the first of the two.
I don't know why Doug didn't apply this patch, but I did apply
his other one.

Looking at the Request.xs code, it seems that we're duping the tempfile's
underlying file descriptor before handing it off to PerlIO.  We
should flush the stdio buffers *before* doing that, but currently 
I don't think we are.

OTOH, I'd expect a potential data-corruption problem here, not a 
segfault.  Have you got a test case or backtrace for us to look at?

-- 
Joe Schaefer