Bug in current snapshots.

[Karl Günter Wünsch <kgw at cswgmbh dot de>]

Hello,

I have been using the snapshots of egcs and gcc for quite some time.
Since the egcs-19990920 I am experiencing a strange problem under
x86-Linux: Whenever I compile one of my projects that is quite
vulnerable to floating point quirks, I have to enable the optimizer to
get a decent result. Compiled without optimizations (for profiling etc.)
I get unpredictable results from my program. I can only assume that it
has something to do with changes that happened from egcs-19990913 to
egcs-19990920 since I am not seeing this problem under any previous
snapshot or release version.

If further information or testing is necessary to track down this
problem I will try to provide all that is necessary but since the
program I am experiencing the problem with is
a) unfinished buisiness that only makes sense to me (graphical output
requiring lot's of interpretation)
and
b) I haven't found the location of the problem in question since the
calculations are rather complex;
tracking down the problem might be a bit of a needle in a haystack
buisiness. If someone could point out the major changes from 19990913 to
19990920 I might be able to boil down the problem in a smaller more
educational example...

regards
Karl Günter Wünsch