------- Comment #6 from paolo dot carlini at oracle dot com 2009-04-03 21:40 ------- (In reply to comment #4) > (In reply to comment #3) > > Created an attachment (id=17584) --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=17584&action=view) [edit] > > Fix a mistake in operator precedence in bits/random.h (_ShiftMin1) > > > It still doesn't work for 32bit: > > /export/gnu/import/gcc/libstdc++-v3/testsuite/20_util/shared_ptr/thread/default_weaktoshared.cc:93:
HJ, I'm sorry, I do not understand your english: you say "it *still* doesn't work", but this *specific* fail was not present in your initial list. Thus, is Ed's patchlet introducing one *additional* fail? Note, in general Ed's patchlet alone *cannot* *fix* the fails I discussed in my previous message, I outlined what is wrong and the proper fix for those. I just want to understand if it makes things worse or not. Note that so far, that is according to the current mainline status, we *cannot* say to have properly *regressions* because all the tests you mentioned in this audit so far are *new*. Just to be accurate. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=39629
