Hi Salvatore, Salvatore Bonaccorso wrote: > On Thu, Feb 11, 2021 at 11:39:09PM +0100, Axel Beckert wrote: > [...] > > Salvatore, Utkarsh: Will also prepare and test at least patches in Git > > for Buster and Stretch. (Hey, I don't want my mutt screen sessions to > > be killed anymore when reading this thread. ;-) We should then > > probably coordinate 1:1 who does the according stable-security and LTS > > uploads. > > Thanks for all your coordinaton, investigation, work on this!
You're welcome. The deep dive into the code and the discussions were
interesting. And I was affected, albeit more in a
self-fulfilling-prophecy way. ;-)
> Sounds good. I propose to have the potential final patch as well first
> slightly exposed first in unstable for a couple of days (2-3)? to see
> if anybody reports any further problems and only then release an
> update in buster, stretch.
Yeah, that's also my thought.
In contrary to my quick and dirty patch in 4.8.0-4, which I
deliberately uploaded with urgency=low, I uploaded this one with
urgency=medium as I'm way more convinced of that patch, especially
after those manual Unicode regression tests I did. (Still thinking how
I can put these into something run in a test suite.)
So if the release team again sets unblock for all packages already in
unstable but not yet in testing at the time of the softfreeze, it will
likely migrate within 2 days if autopkgtest passes. Otherwise we might
want to thing about an unblock request.
But unfortunately — for the first time since I started them in screen
— autopkgtest failed, even on amd64 where I was running it locally,
too.
The subtest which failed was "6 - Session is gone" in
debian/tests/t/dash-U.t. "-U" is slightly related to the patch (if it
does anything these days), but I haven't seen any screen session not
exiting through my tests, so I just requested a retry of that test —
also because these tests (all written by myself as screen has no
upstream test suite) might contain race conditions as it's IPC after
all. And it seems to have passed.
Also suspicious that it passed after 37 seconds while the failure
needed 1m36s to go through. Indeed looks like the first autopkgtest
was run on a system under load and hence might need an additional
"sleep 1" or so. Currently there is none in that specific test —
which adds to the suspicion that this is a race condition.
Anyway, I will now build new packages for Buster and Stretch for those
my own boxes where I use screen most often to test them there under
production conditions. (I use screen much more often on remote systems
and they tend to have stable instead of unstable. :-)
Regards, Axel
--
,''`. | Axel Beckert <a...@debian.org>, https://people.debian.org/~abe/
: :' : | Debian Developer, ftp.ch.debian.org Admin
`. `' | 4096R: 2517 B724 C5F6 CA99 5329 6E61 2FF9 CD59 6126 16B5
`- | 1024D: F067 EA27 26B9 C3FC 1486 202E C09E 1D89 9593 0EDE
signature.asc
Description: PGP signature
