http://gcc.gnu.org/bugzilla/show_bug.cgi?id=54487
--- Comment #7 from Markus Trippelsdorf <markus at trippelsdorf dot de> 2012-09-05 19:20:18 UTC --- (In reply to comment #6) > Markus, I looked at the gcda file you sent but don't see anything obviously > wrong with it. gcov-dump reports that most of the counts in that file are 0 or > very small, so I am not sure why there were messages about exceeding the > maximal count. Yes, as I wrote in my first message the point of failure always varies (for each bootstrap). The attached cp-demangle.gcda was from a failure similar to the one H.J. reported in Comment 2. I guess fixing H.J.'s issue will fix the other failures, too.