Re: master e87e6a2: Fix unlikely races with GnuTLS, datagrams

[Robert Pluim <rpluim@gmail.com>]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Robert Pluim <rpluim@gmail.com>
>> Date: Wed, 13 Feb 2019 14:09:44 +0100
>> Cc: Paul Eggert <eggert@cs.ucla.edu>
>> 
>> eggert@cs.ucla.edu (Paul Eggert) writes:
>> 
>> > branch: master
>> > commit e87e6a24c49542111e669b7d0f1a412024663f8e
>> > Author: Paul Eggert <eggert@cs.ucla.edu>
>> > Commit: Paul Eggert <eggert@cs.ucla.edu>
>> >
>> >     Fix unlikely races with GnuTLS, datagrams
>> 
>> Paul,
>> 
>> there have been reports on help-gnu-emacs of emacs-26.1 failing when
>> using TLS 1.3 to talk to Google servers, and Iʼve just noticed it here
>> as well (although only in unusual circumstances). emacs-27 doesnʼt
>> have this issue, and cherry-picking this commit to emacs-26 fixes it
>> for me. Perhaps we could apply it there?
>
> It's too invasive and was too recently committed to go to the release
> branch, sorry.

OK. We'll just have to live with it for the moment then (Iʼm not about
to propose to people that they switch off TLS1.3 support, since
they'll forget to turn it back on).

Amusingly I only see it because itʼs triggered by
network-security-level > medium, which must change some timing
somewhere (adding logging or running nsm non-byte-compiled makes it go
away).

Robert