This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: gdb.threads/step-after-sr-lock.exp behaving very badly
- From: Pedro Alves <palves at redhat dot com>
- To: Mark Kettenis <mark dot kettenis at xs4all dot nl>
- Cc: gdb-patches at sourceware dot org
- Date: Tue, 11 Feb 2014 10:45:59 +0000
- Subject: Re: gdb.threads/step-after-sr-lock.exp behaving very badly
- Authentication-results: sourceware.org; auth=none
- References: <201402101710 dot s1AHAF0Q022805 at glazunov dot sibelius dot xs4all dot nl>
On 02/10/2014 05:10 PM, Mark Kettenis wrote:
> Hi Pedro,
>
> For some reason the test ends up doing "kill -SIGUSR1 -1". Which
> makes me rather unhappy as it kills all my processes on the machine...
>
> The test hangs early on on OpenBSD (probably because the test program
> blocks all signals).
Ah, userspace threads... I wrote a patch to the test yesterday that makes
it not do that, by using pthread_kill instead of the host's kill, which I
wanted to do anyway to make the test runnable with remote targets, but,
that made me recall why I hadn't done things that way to begin with. The
issue is that the prepare_to_proceed logic only triggers when the thread
that needs to have a breakpoint stepped over was the last that reported
an event. That's actually a bug. I'm working on a patch for that...
> So get_value times out and returns -1. Since
> the effect is fairly disastrous, I went ahead and patched it up with
> the diff below.
Thanks!
--
Pedro Alves