------- Additional Comments From pcarlini at suse dot de 2005-07-22 01:24 ------- For the moment, I give up: in my opinion, a miscompilation is still the most likely possibility. Meaningless behaviors are taking place: for instance, the testcase passes if I exchange the arguments of swap to (my_other_list, my_list) and of course the implementation is symmetric and nobody changed it for eons.
A last comment about mudflap. Right now, I'm not really able to make sense out of some of its outputs: even a single push_back() to a single list leads to errors?!? Really, it would be great if Janis could hunt for the regression. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=22591