This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH] Improve analysis of racy testcases
- From: Antoine Tremblay <antoine dot tremblay at ericsson dot com>
- To: Pedro Alves <palves at redhat dot com>
- Cc: Antoine Tremblay <antoine dot tremblay at ericsson dot com>, Sergio Durigan Junior <sergiodj at redhat dot com>, GDB Patches <gdb-patches at sourceware dot org>
- Date: Mon, 14 Mar 2016 10:04:05 -0400
- Subject: Re: [PATCH] Improve analysis of racy testcases
- Authentication-results: sourceware.org; auth=none
- References: <87r3gcgm91 dot fsf at redhat dot com> <wwokmvqo1vxn dot fsf at ericsson dot com> <878u22wxgk dot fsf at redhat dot com> <wwokk2l51bra dot fsf at ericsson dot com> <56E6B259 dot 4080007 at redhat dot com>
Pedro Alves writes:
> On 03/14/2016 12:31 PM, Antoine Tremblay wrote:
>
>> Just to let know you that the script worked as expected during my
>> testing and it's quite nice to see. The tests however were more flacky
>> then expected so I can't use the output directly guess I'll need a never
>> ignore this test list... (like base/break.exp). Still it's very helpful.
>
> Could you share a bit more on what you found? I wouldn't expect break.exp
> to be flaky. I think some tests that use gdb_test_stdio are still
> flaky with gdbserver, but break.exp does not seem to use it.
>
It's flaky because of a weird problem I have while running GDBServer on
ARM, sometimes out of nowhere I will get a SIGILL just running a program
on GDBServer, clean of breakpoints or anything just loading the program
in GDBServer and doing a continue can trigger it.
See this thread: https://www.sourceware.org/ml/gdb/2015-11/msg00030.html
So it's not because the test itself is flaky but more the platform in general...
Regards,
Antoine