Hallo,
I have been using egfortran from package
gcc 4.2 for six months and bugs showed here and there.
Recently I had occasion to compile large pieces of chemical
computing software.
The result depended on optimization flags:
-O2  all tests result in never ending jobs
-O1  all tests segmentation fault
-O0  all tests passed but a few.

I built a port of gcc 4.2.4 from a modification of official
openbsd port. Basically I accepted all cleanly applied patches
and rejected the rest. The situation is much better but not 
satisfactory.
-O0 there are still failing tests
-O1 some tests don't crash and are passed, others crash
-O2 some tests never end.

I discovered that some assignments were entirely skipped with
-O1.

So I downloaded the NIST f77 testing suite at 
http://www.fortran-2000.com/ArnaudRecipes/fcvs21_f95.html

and ran it with egfortran -O0 (all tests passed) and with
egfortran -O1 (never ending jobs).

I also tried to compile gcc 4.3.1, more or less with the same
strategy but this also resulted in a never ending compilation while
compiling the c++ system.

Please, any information on 
a. current plans of gcc porting/debugging
b. is this never-ending-job motif across different gcc versions
   an impression of mine or
   b.1 is there a reason why this behaviour shows on OpenBSD in particular
   b.2 where should one hunt for the origin of this kind of bug?

Fulvio

Attachment: gcc-4.2.4.tgz
Description: gcc 4.2.4 port

Reply via email to