[PATCH] Prevent restart of crashing non-Cygwin exe

Christopher Faylor cgf-use-the-mailinglist-please@cygwin.com
Sat Nov 5 18:31:00 GMT 2011

On Sat, Nov 05, 2011 at 06:00:25PM +0100, Christian Franke wrote:
>Christopher Faylor wrote:
>> On Thu, Nov 03, 2011 at 01:07:20PM +0100, Corinna Vinschen wrote:
>>> On Nov  2 20:53, Christian Franke wrote:
>>>> On Jun 24, Corinna Vinschen wrote:
>>>>> Hi Christian,
>>>>> On Jun 23 19:52, Christian Franke wrote:
>>>>>> If a non-Cygwin .exe started from a Cygwin shell window segfaults,
>>>>>> Cygwin restarts the .exe 5 times.
>>>>>> [...l]
>>>>>> 	* sigproc.cc (child_info::sync): Add exit_code to debug
>>>>>> 	message.
>>>>>> 	(child_info::proc_retry): Don't retry on unknown exit_code
>>>>>> 	from non-cygwin programs.
>>>>> This looks ok to me, but cgf should have a say here.  He's on vacation
>>>>> for another week, though.
>>>> Problem can still be reproduced with current CVS.  Patch is still
>>>> valid.
>>> Sorry, I forgot about this patch entirely.  Chris, is that patch ok
>>> with you as well?
>> No, it isn't.  Sorry for not stating this earlier.  The problem that
>> this code was intended to solve was actually a transient exit codes from
>> a non-Cygwin process which began with 0xc...
>> I don't believe that I ever saw STATUS_ACCESS_VIOLATION in any of my
>> testing though so adding that earlier in the switch would fix this
>> particular problem.  I'll do that.
>Works as expected with testcase from my first mail:
>$ i686-w64-mingw32-gcc -o crash-w crash.c
>$ ./crash-w
>A drawback is that non-Cygwin programs crash silently.

Thanks for the patch but I chose to do this in a different way.  I've
checked in a change which should accomplish the task of having bash
report the correct error message while allowing easy future extension.


More information about the Cygwin-patches mailing list