On Wed, 15 Jan 2014 18:20:49 +0100, Mark Wielaard wrote:
> I admit it wasn't immediately obvious to me why we were letting the
> child dump core in this case.
BTW it is a regression since
70c3a53baa06b6cdee6e92bd673c1cf977066bc1
tests: Don't use ptrace detach stopped trick. Raise can return.
Although previously it was not much correct as it left the SIGSTOPped process
to be reparented to PID 1 which somehow correctly discarded it.
> So please at least add a comment here why we are forcefully killing the
> child. But I think a nicer fix would be to just let the child exit
> normally after raise () returns.
The stack is corrupt at this point but OK, it can be done.
Thanks,
Jan
--- Begin Message ---
tests/
2014-01-18 Jan Kratochvil <[email protected]>
Fix false FAILs on testsuite with ulimit -c unlimited.
* backtrace-child.c (sigusr2): Call pthread_exit.
(main): Return, do not call abort.
Signed-off-by: Jan Kratochvil <[email protected]>
---
tests/backtrace-child.c | 8 +++++---
1 file changed, 5 insertions(+), 3 deletions(-)
diff --git a/tests/backtrace-child.c b/tests/backtrace-child.c
index 0db1258..512aa23 100644
--- a/tests/backtrace-child.c
+++ b/tests/backtrace-child.c
@@ -100,7 +100,10 @@ sigusr2 (int signo)
if (! gencore)
{
raise (SIGUSR1);
- /* It should not be reached. */
+ /* Do not return as stack may be invalid due to ptrace-patched PC to the
+ jmp function. */
+ pthread_exit (NULL);
+ /* Not reached. */
abort ();
}
/* Here we dump the core for --gencore. */
@@ -218,6 +221,5 @@ main (int argc UNUSED, char **argv)
pthread_join (thread, NULL);
else
raise (SIGUSR2);
- /* Not reached. */
- abort ();
+ return 0;
}
--
1.8.4.2
--- End Message ---
