I can consistently produce Segmentation Faults and/or ICE's when using gcc-3.0 with the above options on large projects (ie gcc 3 and mainline, glibc CVS mainline).
I have submitted at least one PR to glibc PRlibc/2286(their gnats is down -( ) and one to gcc PR2968 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view&pr=2968&database=gcc The solutions I've been able to use work, but still imply something is broken WRT march=athon && (-Os || -ggdb). I can compile with: march=athlon -O2 -g march=athlon -O2 march=athlon -g march=athlon -Os (sometimes) march=athlon -ggdb (sometimes) march=athlon -Os -ggdb (sometimes) If I change s/athlon/i686/ I see no problems so far. I understand the athon code is new (and mostly/somewhat untested?). I also understand the kernel itself is having problems with 2.4.x and athlon code WRT some VIA chipsets (I have one and innumerable oopses sent to lkml). All of my compiling problems are done with Debian unstable and kernel 2.2.19. Is someone responsible for the athlon code? Is it being released with gcc-3.0 as experimental? Is it possible their is a relationship from these problems to the same ones kernel 2.4.x is having? Is there anything I can do to assist? Duron 800 Epox 8KTA3 -- Gordon Sadler
