| Message ID | 20121216211911.GA17639@bromo.med.uc.edu |
|---|---|
| State | New |
| Headers | show |
On Sun, Dec 16, 2012 at 04:19:11PM -0500, Jack Howarth wrote: > 2012-12-16 Jack Howarth <howarth@bromo.med.uc.edu> > > PR sanitizer/55679 > * g++.dg/asan/interception-test-1.C: Skip on darwin. > * lib/target-supports.exp (check_effective_target_swapcontext): Use > check_no_compiler_messages to test support in ucontext.h. The above two changes were already approved, just check those in. > * c-c++-common/asan/rlimit-mmap-test-1.c: Likewise. This doesn't look right. Either you should add code testing for RLIMIT_AS presence into the setrlimit effective target proc, or add setrlimit_as as another effective target. If we want to have more such tests in the future, we don't want to augment all those tests with dg-skip-if for broken targets. > --- gcc/testsuite/c-c++-common/asan/rlimit-mmap-test-1.c (revision 194538) > +++ gcc/testsuite/c-c++-common/asan/rlimit-mmap-test-1.c (working copy) > @@ -2,6 +2,7 @@ > > /* { dg-do run { target setrlimit } } */ > /* { dg-skip-if "" { *-*-* } { "*" } { "-O0" } } */ > +// { dg-skip-if "Darwin lacks RLIMIT_AS in setrlimit" { *-*-darwin* } } > /* { dg-shouldfail "asan" } */ > > #include <stdlib.h> Jakub
On Mon, Dec 17, 2012 at 03:42:59PM +0100, Jakub Jelinek wrote: > On Sun, Dec 16, 2012 at 04:19:11PM -0500, Jack Howarth wrote: > > 2012-12-16 Jack Howarth <howarth@bromo.med.uc.edu> > > > > PR sanitizer/55679 > > * g++.dg/asan/interception-test-1.C: Skip on darwin. > > * lib/target-supports.exp (check_effective_target_swapcontext): Use > > check_no_compiler_messages to test support in ucontext.h. > > The above two changes were already approved, just check those in. > > > * c-c++-common/asan/rlimit-mmap-test-1.c: Likewise. > > This doesn't look right. Either you should add code testing for > RLIMIT_AS presence into the setrlimit effective target proc, or > add setrlimit_as as another effective target. If we want to have more such > tests in the future, we don't want to augment all those tests with > dg-skip-if for broken targets. Jakub, On darwin in /usr/include/sys/resource.h, we have... #define RLIMIT_AS 5 /* address space (resident set size) */ #if !defined(_POSIX_C_SOURCE) || defined(_DARWIN_C_SOURCE) #define RLIMIT_RSS RLIMIT_AS /* source compatibility alias */ so I guess we could craft a setrlimit effective target proc which checks if RLIMIT_AS has the same rlimit ID as RLIMIT_RSS. Any suggestions on how to implement such a check? Jack > > > --- gcc/testsuite/c-c++-common/asan/rlimit-mmap-test-1.c (revision 194538) > > +++ gcc/testsuite/c-c++-common/asan/rlimit-mmap-test-1.c (working copy) > > @@ -2,6 +2,7 @@ > > > > /* { dg-do run { target setrlimit } } */ > > /* { dg-skip-if "" { *-*-* } { "*" } { "-O0" } } */ > > +// { dg-skip-if "Darwin lacks RLIMIT_AS in setrlimit" { *-*-darwin* } } > > /* { dg-shouldfail "asan" } */ > > > > #include <stdlib.h> > > > Jakub
On Mon, Dec 17, 2012 at 10:13:25AM -0500, Jack Howarth wrote: > On darwin in /usr/include/sys/resource.h, we have... > > #define RLIMIT_AS 5 /* address space (resident set size) */ > #if !defined(_POSIX_C_SOURCE) || defined(_DARWIN_C_SOURCE) > #define RLIMIT_RSS RLIMIT_AS /* source compatibility alias */ > > so I guess we could craft a setrlimit effective target proc which checks > if RLIMIT_AS has the same rlimit ID as RLIMIT_RSS. Any suggestions on how > to implement such a check? Or perhaps just return 0 early for [ target *-*-darwin* ] in the setrlimit effective target test, with a comment that it violates POSIX. Jakub
Index: gcc/testsuite/lib/target-supports.exp =================================================================== --- gcc/testsuite/lib/target-supports.exp (revision 194538) +++ gcc/testsuite/lib/target-supports.exp (working copy) @@ -736,7 +736,14 @@ proc check_effective_target_setrlimit {} # Return 1 if the target supports swapcontext, 0 otherwise. proc check_effective_target_swapcontext {} { - return [check_function_available "swapcontext"] + return [check_no_compiler_messages swapcontext executable { + #include <ucontext.h> + int main (void) + { + ucontext_t orig_context,child_context; + if (swapcontext(&child_context, &orig_context) < 0) { } + } + }] } # Return 1 if compilation with -pthread is error-free for trivial Index: gcc/testsuite/g++.dg/asan/interception-test-1.C =================================================================== --- gcc/testsuite/g++.dg/asan/interception-test-1.C (revision 194538) +++ gcc/testsuite/g++.dg/asan/interception-test-1.C (working copy) @@ -3,6 +3,7 @@ // { dg-do run } // { dg-options "-fno-builtin-malloc -fno-builtin-free" } // { dg-shouldfail "asan" } +// { dg-skip-if "Darwin uses mac function interposition" { *-*-darwin* } } #include <stdlib.h> #include <stdio.h> Index: gcc/testsuite/c-c++-common/asan/rlimit-mmap-test-1.c =================================================================== --- gcc/testsuite/c-c++-common/asan/rlimit-mmap-test-1.c (revision 194538) +++ gcc/testsuite/c-c++-common/asan/rlimit-mmap-test-1.c (working copy) @@ -2,6 +2,7 @@ /* { dg-do run { target setrlimit } } */ /* { dg-skip-if "" { *-*-* } { "*" } { "-O0" } } */ +// { dg-skip-if "Darwin lacks RLIMIT_AS in setrlimit" { *-*-darwin* } } /* { dg-shouldfail "asan" } */ #include <stdlib.h>
The attached revised patch disables three asan tests introduced in r194458 which are invalid on darwin. The g++.dg/asan/interception-test-1.C test should be dg-skip-if'd on darwin since mac function interposition is used instead of interception. The c-c++-common/asan/swapcontext-test-1.c test is invalid as ucontext support was deprecated in darwin10 and removed completely in darwin11. The current implementation of check_effective_target_swapcontext in lib/target-supports.exp doesn't catch the deprecation of ucontext support on darwin because it doesn't try to parse the ucontext.h include file. The attached patch changes the check_effective_target_swapcontext to use check_no_compiler_messages with compilation of ucontext.h. It also skips the c-c++-common/asan/rlimit-mmap-test-1.c on darwin since that target lacks RLIMIT_AS in setrlimit. Tested without regression on both x86_64-apple-darwin10 and x86_64-unknown-linux-gnu. Okay for gcc trunk? Jack ps This patch leaves the failures in c-c++-common/asan/global-overflow-1.c and c-c++-common/asan/stack-overflow-1.c as the only unresolved parts of PR55679 on darwin. gcc/testsuite/ 2012-12-16 Jack Howarth <howarth@bromo.med.uc.edu> PR sanitizer/55679 * g++.dg/asan/interception-test-1.C: Skip on darwin. * c-c++-common/asan/rlimit-mmap-test-1.c: Likewise. * lib/target-supports.exp (check_effective_target_swapcontext): Use check_no_compiler_messages to test support in ucontext.h.