At Fri, 12 Apr 2013 14:28:02 +0800, Liu Yuan wrote: > > On 04/12/2013 02:21 PM, MORITA Kazutaka wrote: > > It means that the default handler for the signal couldn't terminate > > the process even though we sent the signal just before the line. If > > we get the message, something goes wrong. > > we are already at crash_handler. doesn't it mean the process is going to > be terminated? What the patch does is to raise the signal manually. I'm > afraid I still don't get it the intention to call signal handler inside > crash_handler().
If we don't call the default handler by re-raising the signal, the process doesn't dump a core file. Actually, current collie and logger cannot generate a core file. Sheep can generate a core file on my environment because the default handler is called after crash_handler(). However, the behavior of the process when we don't exit in the SIGSEGV handler is not defined according to POSIX. If we call exit() in crash_handler(), we cannot get the core file. Thanks, Kazutaka -- sheepdog mailing list [email protected] http://lists.wpkg.org/mailman/listinfo/sheepdog
