"Andreas Karlsson" <andreas.karlsson@xxxxxxxxxxx> writes: >>This is most likely a bug. Unfortunately, this type of bug can be >>somewhat painful to pin down. There are a number of possible causes. >>One possible cause is the problem identified in >>http://gcc.gnu.org/ml/gcc-patches/2007-08/msg01252.html > > I've tried the solution mentioned above and it makes no difference, any > other suggestion or should we accept the fact and call it an (un)harmful > feature? Personal I would prefer to get it byte for byte identical. It's a bug, not a feature. I don't have any further suggestions other than debugging. The first debugging step would be to produce all the dump files and compare them to see where the first difference occurs (use -fdump-unnumbered to remove irrelevant addresses). That should let you identify which pass causes the difference. Unfortunately after that it is just a matter of debugging to find out why the difference occurs. I would encourage you to file a bug report with precise and complete steps for how to recreate the problem. See http://gcc.gnu.org/bugs.html > Btw, the solution wasn't implemented in the gcc-4.2.2 and the ChangeLog > didn't mention why or why not. The bug was identified after gcc 4.2.1 was released, and was not considered serious enough to be fixed on a release branch. After doing a release bug fixes on the release branch are restricted to regressions from past releases or generation of invalid code. Ian
