Let's do an update of the known reported bugs and for trans-mem/libitm.
Thanks to everybody who helped in reporting, fixing, improving and
reviewing trans-mem things.
*Known bugs and eventually fixed in 4.7 or 4.8*
trans-mem:
* Bug 52141 - [trans-mem] ICE due to asm statement in
trans-mem.c:expand_block_tm
* Bug 51173 - XFAIL: libitm.c++/static_ctor.C
error: unsafe function call ‘void __cxa_guard_release(long long
int*)’ within ‘transaction_safe’ function
error: unsafe function call ‘int __cxa_guard_acquire(long long int*)’
within ‘transaction_safe’ function
The static initialization here is using locks. One way to fix that is
to transform the acq/rel always by a transaction.
Making them transaction_pure is not enough because in the
constructor, the transaction may abort and so the lock will not be released.
* Bug 52142 - [trans-mem] inlined transaction_pure functions are
instrumented
* Bug 51752 - trans-mem: publication safety violated
* Bug 51252 - FAIL: c-c++-common/tm/freq.c (internal compiler error)
-> Not fixed for all arch.
* Bug 52042 - libitm.c/clone-1.c (all darwin) and libitm.c/stackundo.c
(darwin 11) execution tests fail with -fpie
for clone-1.c, the problem is that getsectdata doesn't work with
-fpie. Fix proposed but I guess it needs some work to deal with all
situations.
* (No PR) - eh-1.C execution failures seen at -m32/-m64 on x86_64
darwin10/11 when built with Xcode 4.2(.1).
The problem is due to a bug in darwin/Xcode 4.0 to 4.3 with weak
references.
http://gcc.gnu.org/ml/gcc-patches/2012-01/msg00329.html
http://gcc.gnu.org/ml/gcc-patches/2012-01/msg00285.html
http://gcc.gnu.org/ml/gcc-patches/2012-01/msg01676.html
So it fails to detect GCC_CHECK_ELF_STYLE_WEAKREF. Probably needs a
workaround?
*Features for 4.8*
trans-mem:
* Bug 51771 - trans-mem: abnormal edges get lost or corrupted
-> Solved with return_twice for 4.7 but needs work for 4.8.
-> part of the fix is here:
http://gcc.gnu.org/ml/gcc/2012-01/msg00099.html
* Bug 46480 - [trans-mem] Uninstrumented code path is missing
libitm:
* Bug 51855 - improve calculation of stack bottom in libitm's undolog
* Bug 51169 - XFAIL: libitm.c/dropref-2.c execution test
Bug 51170 - XFAIL: libitm.c/dropref.c execution test
Bug 51172 - XFAIL: libitm.c++/dropref.C execution test
Same problem for all the 3 PR, ie missing ABI feature.
ABI feature will be implemented eventually.
Do tests shoud be removed and added later when the feature is
implemented?
* (No PR) [libitm] libitm.h is not installed
The fix is simple but maybe a reason for not exposing this to TM user?
By the way, should it be renamed to simply itm.h?
Thanks!
--
Patrick Marlier.
