[sheepdog] [PATCH] call a default signal handler in crash_handler()

Liu Yuan namei.unix at gmail.com
Fri Apr 12 11:14:17 CEST 2013


On 04/12/2013 04:30 PM, MORITA Kazutaka wrote:
> I didn't consider that the unexpected signal triggers the
> crash_handler.  Then the reason we reach the line are:
> 
>  - The default handler (for SIGSEGV, SIGABRT, SIGBUS, SIGILL, or
>    SIGFPE) doesn't terminate the process unexpectedly.
> 
>  - The unexpected signal has come and its default handler doesn't
>    terminate the process.
> 
> Is it okay?  Then I think the current message shows what has happened
> clearly.  In either case, we should stop the process immediately
> because what we cannot expect has happened.
> 
> Anyway, we shouldn't reach the line.  Is it better to print "something
> goes wrong" simply if the current message is confusing?

Okay, I didn't get clear behavior of install_handler(), which will treat
the signal that happens twice differently. I think we should explicitly
add the explanation of this behavior for it.

Then the message for sd_printf should be better rephrased as "default
handler of the re-raised signal xxx didn't work expectedly".

Thanks,
Yuan



More information about the sheepdog mailing list