emacs freezes

emacs freezes

[Jose A. Ortega Ruiz]

Hi. Since some time ago, my emacs freezes intermitently. That is, it
ceases to react to any keystroke and all frames become blank. There is
no CPU activity (top lists the emacs process as stopped). The only way
out is kill. If i attach strace to the process instance i got the
following trace continously:

Process 4067 attached - interrupt to quit
futex(0x407d9860, FUTEX_WAIT, 2, NULL)  = -1 EINTR (Interrupted system call)
--- SIGIO (I/O possible) @ 0 (0) ---
sigreturn()                             = ? (mask now [SEGV])

so it seems some kind of problem with the 'futex' syscall. I'm using
linux 2.6.8 (with the futex kernel config flag enabled) on a debian
sid. window manager is usually ion (but, iirc, it also happens with
pekwm and sawfish), and the freeze (almost?) always happen when
switching between emacs frames. i usually have separate emacs frames
with eshell, gnus and w3m, in case that matters.

i've tried both emacs 21.3.1 (the stock debian package) and the latest
cvs (21.3.50). same problem in both, but when using cvs + gtk2.0 the
frequency of the freeze increases dramatically from approx. once a day
(with 21.3.1 and xawd emacs cvs) to once every few minutes (with gtk).

any help would be greatly appreciated, since i spent 90% of my
computer time inside emacs :)

thanks,
jao
-- 
We shall do a much better programming job, provided we approach the
task with a full appreciation of its tremendous difficulty, provided
that we respect the intrinsic limitations of the human mind and
approach the task as very humble programmers. (Alan Turing)

Re: emacs freezes

[Stefan Monnier]

> Process 4067 attached - interrupt to quit
> futex(0x407d9860, FUTEX_WAIT, 2, NULL)  = -1 EINTR (Interrupted system call)
> --- SIGIO (I/O possible) @ 0 (0) ---
> sigreturn()                             = ? (mask now [SEGV])

> so it seems some kind of problem with the 'futex' syscall. I'm using
> linux 2.6.8 (with the futex kernel config flag enabled) on a debian

Could you explain what is "futex" ?

> i've tried both emacs 21.3.1 (the stock debian package) and the latest
> cvs (21.3.50). same problem in both, but when using cvs + gtk2.0 the
> frequency of the freeze increases dramatically from approx. once a day
> (with 21.3.1 and xawd emacs cvs) to once every few minutes (with gtk).

Better focus on the 21.3.50 version (the only one that might get fixed
since the other is already released).  Send a detailed bug report to
emacs-pretest-bug@gnu.org (ideally via M-x report-emacs-bug).


        Stefan

Re: emacs freezes

[Jose A. Ortega Ruiz]

Stefan Monnier <monnier@iro.umontreal.ca> writes:

>> Process 4067 attached - interrupt to quit
>> futex(0x407d9860, FUTEX_WAIT, 2, NULL)  = -1 EINTR (Interrupted system call)
>> --- SIGIO (I/O possible) @ 0 (0) ---
>> sigreturn()                             = ? (mask now [SEGV])
>
>> so it seems some kind of problem with the 'futex' syscall. I'm using
>> linux 2.6.8 (with the futex kernel config flag enabled) on a debian
>
> Could you explain what is "futex" ?
>

i'm not an expert, but they're a new (as of 2.6 kernels) low level
syncronisation primitive that work in user space and is used to
construct higher level abstractions such as mutexes or semaphores. i
guess they're used internally in calls like select. from the man page,
an EINTR during the system call above is provoked by "signals or other
spurious wakeups"... not that i can make much sense of it :(

>
> Better focus on the 21.3.50 version (the only one that might get fixed
> since the other is already released).  Send a detailed bug report to
> emacs-pretest-bug@gnu.org (ideally via M-x report-emacs-bug).
>

that's the version i'm use regularly. bug report sent. thank you.

jao
-- 
"Beware of bugs in the above code; I have only proved it correct,
not tried it." -- Donald Knuth (in a memo to Peter van Emde Boas)