Bug#507817: screen -r still sometimes fails with "WriteMessage: Bad file descriptor"
Sami Liedes
Version: 4.0.3-11
Severity: normal
Hi,
I still sometimes get the error "WriteMessage: Bad file descriptor"
when running screen -r with 4.0.3-11. This is obviously related to
#477739, however the conditions seem somewhat different so I decided
to open a different bug.
Contrary to #477739, I have two sessions:
------------------------------------------------------------
sliedes@lh:~/scratch/chess/scid-open$ screen -ls
There are screens on:
25929.pts-1.lh (11/27/2008 06:47:30 AM) (Detached)
5942.pts-4.lh (11/26/2008 02:38:47 AM) (Attached)
2 Sockets in /var/run/screen/S-sliedes.
------------------------------------------------------------
At least when I last time saw it happen, that was the state of the
sessions: One (the one to be resumed) was detached, the other
attached, and I did not use -d. I'm not exactly sure if it has been so
always.
Unfortunately this seems to happen rather rarely (how rarely? I don't
know, I tried screen -r 27 times and it happened on times #1 and #27
now). I've seen it maybe 10-20 times now. Also I couldn't reproduce it
under strace with maybe 200 tries.
Please tell me if you need me to further try to debug it somehow (you
can assume I know how to use strace, gdb and friends, I'm a
developer).
Sami
-- System Information:
Debian Release: lenny/sid
APT prefers unstable
APT policy: (500, 'unstable'), (500, 'testing')
Architecture: amd64 (x86_64)
Kernel: Linux 2.6.27.7 (SMP w/4 CPU cores; PREEMPT)
Locale: LANG=en_US.UTF-8, LC_CTYPE=en_US.UTF-8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/bash
Versions of packages screen depends on:
ii libc6 2.7-16 GNU C Library: Shared libraries
ii libncursesw5 5.7+20081129-1 shared libraries for terminal hand
ii libpam0g 1.0.1-4+b1 Pluggable Authentication Modules l
screen recommends no packages.
screen suggests no packages.
-- no debconf information
--
To UNSUBSCRIBE, email to debian-bugs-...@lists.debian.org
with a subject of "unsubscribe". Trouble? Contact listm...@lists.debian.org
Jan Christoph Nordholz
sorry it took me so long to get back to you.
> I still sometimes get the error "WriteMessage: Bad file descriptor"
> when running screen -r with 4.0.3-11. This is obviously related to
> #477739, however the conditions seem somewhat different so I decided
> to open a different bug.
Hm. I haven't seen this bug for quite some time now, so any further
material (gdb dumps, strace logs) is highly appreciated. I'll have a
look at the code and check whether I've missed another race condition
(or failed to properly fix the one from #477739).
Thanks!
Jan
Sami Liedes
> > I still sometimes get the error "WriteMessage: Bad file descriptor"
> > when running screen -r with 4.0.3-11. This is obviously related to
> > #477739, however the conditions seem somewhat different so I decided
> > to open a different bug.
>
> Hm. I haven't seen this bug for quite some time now, so any further
> material (gdb dumps, strace logs) is highly appreciated. I'll have a
> look at the code and check whether I've missed another race condition
> (or failed to properly fix the one from #477739).
Now that I think, I haven't seen it for a while either. And thinking
further, it might have been that the screen I tried to attach to
contained lots of screens started with :screen /dev/pts/nn (qemu with
a serial console). Although for some reason I failed to mention that
in my report, so it might be that either I ruled that out or forgot --
hard to remember which :(
Sami
Vincent Lefevre
On 2008-12-04 19:50:35 +0200, Sami Liedes wrote:
> I still sometimes get the error "WriteMessage: Bad file descriptor"
I also get this error with a probability of about a few percents.
Note: I usually don't see this error directly because I invoke screen
via a script from my window manager, and the terminal window just
immediately disappears without I have the time to see anything in it
(I'll have to update my script so that error messages can be visible).
As I was wondering, I tried screen manually:
zira:~> =screen -T screen-bce -dRR mutt -t mutt
Remove dead screens with 'screen -wipe'.
WriteMessage: Bad file descriptor
(The = in zsh allows to bypass any alias and shell function.)
--
Vincent Lefèvre <vin...@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
Vincent Lefevre
> zira:~> =screen -T screen-bce -dRR mutt -t mutt
> Remove dead screens with 'screen -wipe'.
> WriteMessage: Bad file descriptor
is not detached, i.e. this error occurs quite early.
I could not reproduce the problem with strace.
Axel Beckert
Vincent Lefevre wrote:
> On 2015-07-30 14:20:13 +0200, Vincent Lefevre wrote:
> > zira:~> =screen -T screen-bce -dRR mutt -t mutt
> > Remove dead screens with 'screen -wipe'.
> > WriteMessage: Bad file descriptor
>
> When this happens while a session is already attached, the session
> is not detached, i.e. this error occurs quite early.
upstream bug report where this error message appears) suggests that
this issue maybe related to to session creation time patch in Debian:
https://sources.debian.net/src/screen/4.3.1-1/debian/patches/80EXP_session_creation_time.patch/
Any chance that you can rebuild Debian's screen without(*) that patch
and check if you get the same percentage of error messages?
(*) It's the last patch so doing "quilt pop -a" and then removing it's
entry from debian/patches/series before rebuilding the package
should suffice.
Regards, Axel
--
,''`. | Axel Beckert <a...@debian.org>, http://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
Vincent Lefevre
On 2015-07-30 15:51:43 +0200, Axel Beckert wrote:
> Any chance that you can rebuild Debian's screen without(*) that patch
> and check if you get the same percentage of error messages?
official package) while I didn't change anything.
--
Vincent Lefèvre <vin...@vinc17.net> - Web: <https://www.vinc17.net/>
100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/>
Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
Vincent Lefevre
> upstream bug report where this error message appears) suggests that
> this issue maybe related to to session creation time patch in Debian:
> https://sources.debian.net/src/screen/4.3.1-1/debian/patches/80EXP_session_creation_time.patch/
>
> Any chance that you can rebuild Debian's screen without(*) that patch
> and check if you get the same percentage of error messages?
>
> (*) It's the last patch so doing "quilt pop -a" and then removing it's
> entry from debian/patches/series before rebuilding the package
> should suffice.
So, I rebuilt screen without this patch on September 21, and
since then, the problem has no longer occurred.
Note: of course, I have not removed the dead screens.
Vincent Lefevre
Control: tags -1 patch
I'm tagging patch since the patch is to reverse the patch as mentioned
below.
On 2015-10-17 15:15:22 +0200, Vincent Lefevre wrote:
> On 2015-07-30 15:51:43 +0200, Axel Beckert wrote:
> > https://savannah.gnu.org/bugs/?func=detailitem&item_id=43744 (the only
> > upstream bug report where this error message appears) suggests that
> > this issue maybe related to to session creation time patch in Debian:
> > https://sources.debian.net/src/screen/4.3.1-1/debian/patches/80EXP_session_creation_time.patch/
> >
> > Any chance that you can rebuild Debian's screen without(*) that patch
> > and check if you get the same percentage of error messages?
> >
> > (*) It's the last patch so doing "quilt pop -a" and then removing it's
> > entry from debian/patches/series before rebuilding the package
> > should suffice.
>
> Several weeks ago, the problem was occurring again quite often.
> So, I rebuilt screen without this patch on September 21, and
> since then, the problem has no longer occurred.
the culprit patch, and the bug occurs again:
Remove dead screens with 'screen -wipe'.
WriteMessage: Bad file descriptor
Now this occurs very often:
zira:~> screen -dRR <15:22:36
WriteMessage: Bad file descriptor
zira:~[1]> screen -dRR <15:22:39
WriteMessage: Bad file descriptor
zira:~[1]> screen -dRR <15:22:40
zira:~> screen -dRR <15:22:42
WriteMessage: Bad file descriptor
zira:~[1]> screen -dRR <15:22:44
zira:~> screen -dRR <15:22:45
zira:~> screen -dRR <15:22:47
WriteMessage: Bad file descriptor
zira:~[1]> screen -dRR <15:23:00
zira:~> screen -dRR <15:23:04
WriteMessage: Bad file descriptor
zira:~[1]> screen -dRR <15:23:05
WriteMessage: Bad file descriptor
zira:~[1]> screen -dRR <15:23:06
WriteMessage: Bad file descriptor
zira:~[1]> screen -dRR <15:23:07
Vincent Lefevre
> Now this occurs very often:
But I cannot reproduce the failure with strace.
A race condition somewhere?
Axel Beckert
Vincent Lefevre wrote:
> > Now this occurs very often:
> [...]
>
> But I cannot reproduce the failure with strace.
> A race condition somewhere?
issues reproducing it as well as why users see it with some versions
more often than with others despite the accused patch hasn't seen a
change for quite a while.
I'll check the patch for a potential race condition. But I do also
consider disabling it as there nowadays are enough evidentiary facts
that it's really the cause for this issue.
My preference is neverthelesss to keep the patch -- especially if the
patch can be fixed.
Axel Beckert
Axel Beckert wrote:
> Vincent Lefevre wrote:
> > On 2016-06-24 15:23:52 +0200, Vincent Lefevre wrote:
> > > Now this occurs very often:
> > [...]
> >
> > But I cannot reproduce the failure with strace.
> > A race condition somewhere?
>
> Thanks for that hint. That is a possible explaination why I have
> issues reproducing it as well as why users see it with some versions
> more often than with others despite the accused patch hasn't seen a
> change for quite a while.
calls) with one local build, I was later no more able to do so despite
having built the package from the very same commit again. Also
installing 4.4.0-3 didn't help to reproduce the issue again, despite
Vincent confirmed that it happened with that version on his machine.
:-/
> I'll check the patch for a potential race condition. But I do also
> consider disabling it as there nowadays are enough evidentiary facts
> that it's really the cause for this issue.
the issue, but since he couldn't reproduce it nor I can't reproduce it
anymore, even with the old packages, I can not verify or guarantee
that the fix actually works for all users.
So please reopen this bug report in case you run into the "WriteMessage:
Bad file descriptor" again with Debian's screen package, version
4.4.0-4 or later.
I'd also be happy about confirmations that this bug is indeed fixed
with today's upload (4.4.0-4).
Thanks in advance for any feedback on this issue!