|
From: Bart V. A. <bar...@gm...> - 2008-02-26 19:36:48
|
Hello Julian,
The result of the regression test helgrind/tests/tc20_verifywrap
depends on the platform it is executed on (asserts with glibc 2.7 but
not with older glibc versions). Is it OK if I fix this by committing
the patch below ?
Index: helgrind/tests/tc20_verifywrap.c
===================================================================
--- helgrind/tests/tc20_verifywrap.c (revision 7483)
+++ helgrind/tests/tc20_verifywrap.c (working copy)
@@ -241,8 +241,8 @@
memset(&s1, 0x55, sizeof(s1));
r= sem_wait(&s1); /* assert(r != 0); */
- /* this really ought to fail, but it doesn't. */
- r= sem_post(&s1); assert(!r);
+ /* this only fails with glibc 2.7 or later. */
+ r= sem_post(&s1);
fprintf(stderr, "\nFIXME: can't figure out how to verify wrap of "
"sem_post\n\n");
Bart.
|
|
From: Julian S. <js...@ac...> - 2008-02-26 19:54:52
|
On Tuesday 26 February 2008 20:36, Bart Van Assche wrote: > Hello Julian, > > The result of the regression test helgrind/tests/tc20_verifywrap > depends on the platform it is executed on (asserts with glibc 2.7 but > not with older glibc versions). Is it OK if I fix this by committing > the patch below ? What is the effect on platforms < glibc 2.7? Does it cause the test to fail on those? J > > > Index: helgrind/tests/tc20_verifywrap.c > =================================================================== > --- helgrind/tests/tc20_verifywrap.c (revision 7483) > +++ helgrind/tests/tc20_verifywrap.c (working copy) > @@ -241,8 +241,8 @@ > memset(&s1, 0x55, sizeof(s1)); > r= sem_wait(&s1); /* assert(r != 0); */ > > - /* this really ought to fail, but it doesn't. */ > - r= sem_post(&s1); assert(!r); > + /* this only fails with glibc 2.7 or later. */ > + r= sem_post(&s1); > fprintf(stderr, "\nFIXME: can't figure out how to verify wrap of " > "sem_post\n\n"); > > > Bart. |
|
From: Bart V. A. <bar...@gm...> - 2008-02-27 07:18:38
|
On Tue, Feb 26, 2008 at 8:51 PM, Julian Seward <js...@ac...> wrote: > > On Tuesday 26 February 2008 20:36, Bart Van Assche wrote: > > The result of the regression test helgrind/tests/tc20_verifywrap > > depends on the platform it is executed on (asserts with glibc 2.7 but > > not with older glibc versions). Is it OK if I fix this by committing > > the patch below ? > > What is the effect on platforms < glibc 2.7? Does it cause the > test to fail on those? The patch has no effect on platforms with glibc version < 2.7 (regression test still succeeds), and the patch makes the tc20_verifywrap regression test succeed on platforms with glibc version >= 2.7. Or: the patch ensures that the behavior of this regression test does no longer depend on the glibc version. Bart. |
|
From: Julian S. <js...@ac...> - 2008-02-27 10:56:53
|
On Wednesday 27 February 2008 08:18, Bart Van Assche wrote: > On Tue, Feb 26, 2008 at 8:51 PM, Julian Seward <js...@ac...> wrote: > > On Tuesday 26 February 2008 20:36, Bart Van Assche wrote: > > > The result of the regression test helgrind/tests/tc20_verifywrap > > > depends on the platform it is executed on (asserts with glibc 2.7 but > > > not with older glibc versions). Is it OK if I fix this by committing > > > the patch below ? > > > > What is the effect on platforms < glibc 2.7? Does it cause the > > test to fail on those? > > The patch has no effect on platforms with glibc version < 2.7 > (regression test still succeeds), and the patch makes the > tc20_verifywrap regression test succeed on platforms with glibc > version >= 2.7. Or: the patch ensures that the behavior of this > regression test does no longer depend on the glibc version. Ok; commit. Thanks, J |