[vlc-commits] Restore correct SIGCHLD handling
Rémi Denis-Courmont
git at videolan.org
Sun Jan 30 13:21:02 CET 2011
vlc | branch: master | Rémi Denis-Courmont <remi at remlab.net> | Sun Jan 30 14:20:26 2011 +0200| [5dc560308c06e1862a0250651e5595fc7b0c2a82] | committer: Rémi Denis-Courmont
Restore correct SIGCHLD handling
This affects way too many code paths. Let KDE fix their own crap.
> http://git.videolan.org/gitweb.cgi/vlc.git/?a=commit;h=5dc560308c06e1862a0250651e5595fc7b0c2a82
---
bin/vlc.c | 11 +++++------
1 files changed, 5 insertions(+), 6 deletions(-)
diff --git a/bin/vlc.c b/bin/vlc.c
index 1bc6a27..00da32c 100644
--- a/bin/vlc.c
+++ b/bin/vlc.c
@@ -157,12 +157,11 @@ int main( int i_argc, const char *ppsz_argv[] )
sigaddset (&set, SIGPIPE);
/* SIGCHLD must be dequeued to clean up zombie child processes.
- * Furthermore the handler must not be set to SIG_IGN (see above). */
- /* Unfortunately, the QProcess class from Qt4 has a bug. It installs a
- * custom signal handlers and gets stuck if it is not called. So we cannot
- * use sigwait() for SIGCHLD:
- * http://bugs.kde.org/show_bug.cgi?id=260719 */
- //sigaddset (&set, SIGCHLD);
+ * Furthermore the handler must not be set to SIG_IGN (see above).
+ * We cannot pragmatically handle EINTR, short reads and short writes
+ * in every code paths (including underlying libraries). So we just
+ * block SIGCHLD in all threads, and dequeue it with sigwait() below. */
+ sigaddset (&set, SIGCHLD);
#ifdef HAVE_MAEMO
sigaddset (&set, SIGRTMIN);
More information about the vlc-commits
mailing list