httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Greg Ames <grega...@apache.org>
Subject Re: cvs commit: httpd-2.0/modules/http http_request.c
Date Thu, 20 Jun 2002 20:08:38 GMT
Justin Erenkrantz wrote:
> 
> On Thu, Jun 20, 2002 at 03:41:51PM -0400, Cliff Woolley wrote:
> >
> > Somebody remind me why that ap_status_drops_conn call was added in the
> > first place?  It was a recent change, and seemed important at the time.
> 
> Woah, woah, woah.  Please revert.
> 
> If the status results in a dropped connection, we can't read the
> body because that status code means we don't trust the input.  This
> also means that the r->status code can not be the recursive code (which
> doesn't result in a dropped connection), but must remain the error
> that we just got in ap_die, so that we know that we're not supposed
> to read the body.  This is central to our 413 fixes.

We cannot tolerate infinite recursion if an ErrorDocument gets an error.  The if
statement I changed is how we detect this sort of recursive error.  Please take
a look at the backtrace I posted to dev@httpd.apache.org, under the subject
"2.0.38-39 lockup problem ?".  

If we need to prevent reading the body for certain types of errors, we can't do
it at the expense of loosing our recursive error protection.

Greg

Mime
View raw message