This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH v4 00/18] All-stop on top of non-stop
- From: Joel Brobecker <brobecker at adacore dot com>
- To: Luis Machado <lgustavo at codesourcery dot com>
- Cc: Pedro Alves <palves at redhat dot com>, gdb-patches at sourceware dot org
- Date: Wed, 12 Aug 2015 13:33:22 -0700
- Subject: Re: [PATCH v4 00/18] All-stop on top of non-stop
- Authentication-results: sourceware.org; auth=none
- References: <1432250354-2721-1-git-send-email-palves at redhat dot com> <55C4E3BD dot 8040801 at redhat dot com> <20150812183208 dot GA24901 at adacore dot com> <55CBA0D1 dot 5000203 at codesourcery dot com> <20150812195948 dot GH22245 at adacore dot com> <55CBAA17 dot 4040605 at codesourcery dot com>
I tried to clarify the purpose of the testcase by adding a comment
inside it. Let me know if this isn't enough, and we'll try to improve.
> Ok. So i think i misunderstood the purpose of the testcase there. In reality
> the testcase is not testing the fix itself, but rather introducing a new
> test not related to the problem, except in the amd64 architecture, where it
> really tests the problem.
>
> It just confused me that the test is generic for other non-amd64
> architectures and specific to the amd64 problem you saw.
>
> I'd expect a generic solib test to be included in one of our shared library
> tests, but then you'd have to shape it in a way that would exercise your
> displaced stepping problem.
I understand the first paragraph, but I'm having trouble with the last
one. The testcase as I wrote it does exercise the issue being fixed
on amd64, and I verified that I get 1 FAIL without the patch. Did
I misunderstand you?
> I see your targets are mostly x86. I can give it a try on a few more
> (powerpc, mips, arm, nios) and let you know what i see. How does that sound?
Sounds good to me.
Bare in mind that the list of targets AdaCore tests on already includes
powerpc and sparc, and I forgot ARM that we also test (as a cross,
debugging via GDBserver). So I think it's fairly wide already.
Our testsuite uses a different technology than dejagnu, though, so we
might indeed get some surprises, but I think they will have more to do
with scripting than GDB: Except for the platforms that we don't test
at AdaCore, I'd expect that any error will mostly be about building
the program rather than debugging it.
--
Joel