On 04/29/13 12:09, Ed Maste wrote:
> If my understanding is correct your replacement test case should still
> pass even with the following change:
>
> - sa.sa_flags = SA_RESETHAND | SA_NODEFER;
> + sa.sa_flags = SA_RESETHAND | SA_NODEFER | SA_SIGINFO;
Not exactly, since in that case one also ne
On 25 April 2013 16:23, Paul Eggert wrote:
>
> I guess my test case wasn't a correct simplification of
> the gnulib test case. Could you please try the following
> test case instead? Thanks.
If my understanding is correct your replacement test case should still
pass even with the following chan
On 25 April 2013 22:39, Daiki Ueno wrote:
> I can still reproduce this on 9.1-RELEASE. Perhaps you are missing
> -pthread cflag when compiling?
Argh, you are right. What happened here is I built Paul's first test
using -lpthread, but it doesn't demonstrate the problem. Then I
somehow missed it
Ed Maste writes:
> Tested on 8.4-PRERELEASE, 9.1-STABLE (the 9.x branch after 9.1
> released), and 10.0-CURRENT, all returned 0. Perhaps Daiki Ueno can
> test in the environment that originally demonstrated the problem.
I can still reproduce this on 9.1-RELEASE. Perhaps you are missing
-pthrea
On 25 April 2013 16:23, Paul Eggert wrote:
> On 04/25/13 12:07, Ed Maste wrote:
>> I don't have a 9.1-RELEASE machine handy, but get the expected zero
>> return from this test program on the following FreeBSD versions (one
>> before and one after 9.1-RELEASE):
>
> I guess my test case wasn't a cor
On 04/25/13 12:07, Ed Maste wrote:
> I don't have a 9.1-RELEASE machine handy, but get the expected zero
> return from this test program on the following FreeBSD versions (one
> before and one after 9.1-RELEASE):
I guess my test case wasn't a correct simplification of
the gnulib test case. Could
Hmm, it seems this has been sitting in my drafts folder for quite some time.
> In that case perhaps we should leave the Gnulib test alone,
> and report a bug to the FreeBSD folks. Does the following
> simple test case elicit the bug for you? It should exit with status 0
> on any host conforming
On 01/29/2013 01:31 AM, Daiki Ueno wrote:
> Yes, but from the gettext Hydra build log[1], FreeBSD 8.x does not seem
> to have that behavior. Perhaps it might be a bug in the latest FreeBSD
> kernel.
>
> Footnotes:
> [1] http://hydra.nixos.org/build/3910833/nixlog/1
In that case perhaps we shou
Sorry for the delay.
Paul Eggert writes:
> Always reporting SA_SIGINFO violates POSIX but I doubt whether
> anybody really cares, so does the following patch fix things for you?
> It simply adjusts the test so that it doesn't check for this bug.
Yes, but from the gettext Hydra build log[1], Fre
Thanks for looking into it. From the trace, my guess
is that FreeBSD always reports that SA_SIGINFO is set
because FreeBSD always passes the relevant info to signal handlers
(at the low level, this doesn't cause problems so it's OK).
Always reporting SA_SIGINFO violates POSIX but I doubt whether
a
Paul Eggert writes:
> On 01/17/13 23:04, Daiki Ueno wrote:
>> It seems -fopenmp flag causes the error. It can also be reproduced
>> without openmp, but with CFLAGS="-O2 -pthread".
>
> I can't reproduce it with Fedora 17, but I guess that's not
> too surprising.
Yes.
> Can you send the output o
On 01/17/13 23:04, Daiki Ueno wrote:
> It seems -fopenmp flag causes the error. It can also be reproduced
> without openmp, but with CFLAGS="-O2 -pthread".
I can't reproduce it with Fedora 17, but I guess that's not
too surprising.
Can you send the output of this shell command?
truss ./test-sig
Hi,
On FreeBSD 9.1, I got an assertion failure in test-sigaction.c when the
test case is compiled with:
$ gnulib-tool --create-testdir --dir=test-sa-openmp sigaction openmp
$ cd test-sa-openmp
$ ./configure --enable-openmp
$ make
$ make check
...
test-sigaction.c:69: assertion faile
13 matches
Mail list logo
