threads / discuss / 12375

git-daemon hangs in futex()

Subject: git-daemon hangs in futex()

## tl;dr

2 messages between Feb 28, 2008 and Feb 29, 2008.

replies: 1people: 1as markdown or json

Johan Sørensen· Feb 28, 2008, 22:41 UTC · lore

Hi,

I'm having an issue with my git-daemon (1.5.4.3 on ubuntu gutsy), it works fine for a few minutes, then it hangs (eg, "git clone git://" never comes past "initialized empty git repo" adnd just sits and waits) and strace on the git-daemon says: $ strace -Ff -p 27453 Process 27453 attached - interrupt to quit futex(0x2b270fb50980, FUTEX_WAIT, 2, NULL) = -1 EINTR (Interrupted system call)

.. and nothing more. But, if I attach to the git-daemon process with strace before that happens, git-daemon keeps on working fine, so I can change the outcome by observing it which is always nice, but not too helpful. I don't think this is related to Wincents problem some time ago[1], because I can reproduce this when running dashed/dashless and with/without --detach regardless of my env. Here's my git-daemon args: /usr/local/bin/git --exec-path=/usr/local/bin daemon --base-path=/home/git/repositories --verbose --export-all --pid-file=/home/git/run/git-daemon.pid --detach

I'm not even sure if this is directly related to git or if it's some other library causing the futex wait, any hints would be greatly appreciated.

[1]: http://kerneltrap.org/mailarchive/git/2008/2/5/732214

Thanks JS

Johan Sørensen· Feb 29, 2008, 14:26 UTC · re: Johan Sørensen · lore

Re: git-daemon hangs in futex()

On Thu, Feb 28, 2008 at 11:41 PM, Johan Sørensen <johan@johansorensen.com> wrote:

Show 14 quoted lines

> Hi,
>
>  I'm having an issue with my git-daemon (1.5.4.3 on ubuntu gutsy), it
>  works fine for a few minutes, then it hangs (eg, "git clone git://"
>  never comes past "initialized empty git repo" adnd just sits and
>  waits) and strace on the git-daemon says:
>  $ strace -Ff -p 27453
>  Process 27453 attached - interrupt to quit
>  futex(0x2b270fb50980, FUTEX_WAIT, 2, NULL) = -1 EINTR (Interrupted system call)
>
>  .. and nothing more. But, if I attach to the git-daemon process with
>  strace before that happens, git-daemon keeps on working fine, so I can
>  change the outcome by observing it which is always nice, but not too
>  helpful.

Upon further investigation, it seems that this could be related to me updating glibc to 2.6.1. Google suggest that it can because libc switched to NPTL. I'm not familiar with the internals of it, so I can't decide if this is a git issue (by indirect usage of glibc functions) or not?

Thanks JS