bug#13546: 24.2.92; Error(s) when sending emails

[Eli Zaretskii <eliz@gnu.org>]

> From: Ted Zlatanov <tzz@lifelogs.com>
> Cc: 13546@debbugs.gnu.org,  wxhgmqzgwmuf@spammotel.com
> Date: Fri, 01 Mar 2013 14:40:59 -0500
> 
> On Fri, 01 Mar 2013 16:03:41 +0200 Eli Zaretskii <eliz@gnu.org> wrote: 
> 
> >> From: Ted Zlatanov <tzz@lifelogs.com>
> >> Cc: 13546@debbugs.gnu.org,  wxhgmqzgwmuf@spammotel.com
> >> Date: Fri, 01 Mar 2013 08:36:23 -0500
> >> 
> >> I didn't see this suggested: can you try from another platform besides W32?
> 
> EZ> Unless you are thinking about a bug in GnuTLS, I don't see how this
> EZ> could help.  The problem is quite clearly in the infrastructure used
> EZ> in the w32 build of Emacs.  Any other platform will not have any
> EZ> comparable infrastructure, as they all use system-provided facilities
> EZ> like signals and pty's.
> 
> I'd like to know that there's a working Emacs and a broken one, so we
> can do some differential analysis.

The symptoms we saw up till now (see the discussions) all so
overwhelmingly point towards the w32-specific code that deals with
subprocesses and network connections that it is hard to believe
comparison with any other platform will be of any benefit whatsoever.

In any case, we already have a working Emacs (24.2.91) and a broken
one (24.2.92 and 24.2.93), so if you want to compare them, please do.

> Right now we're crossing our fingers that the latest pretest works.

_I_ am crossing fingers.  Because I see no changes between 24.2.91 and
the current emacs-24 branch that are specific to w32 and touch the
relevant areas of Emacs.  If the current release branch is still
buggy, I simply don't know what else could be the reason, and will
advise Glenn to proceed with the release candidate, this problem
notwithstanding.

> Also, if we can determine definitely that the problem only affects W32
> as we have supposed so far, it makes the debugging easier and the
> problem less of a release blocker.

No one else reported anything similar, not on Windows, not elsewhere.

> I brought up GnuTLS because it's involved in some of Sebastian's setup,
> and it has special glue in Emacs for W32.  As with my other suggestions,
> I'm just trying to do some differential analysis to see if there's some
> kind of bisection that will reduce the set of possible causes.  But I
> agree with you that it's an unlikely cause... so I'll be quiet now :)

No need to be quiet.  Any comments or thoughts on this are greatly
appreciated.