Re: egcs-1.1 segfaults!
On Tue, 22 Sep 1998, Michele Comitini wrote:
> No egcs fault this time!
> I debugged the thing and I discovered it was the ncurses lib I used,
> I switched to the newest and now it works.
That helps. Yeah, there were some really buggy ncurses out there before
:-) And the battle rages on (see the devel list on the versions
argument...it's getting strange).
> Yes compiling with just -O works fine. Maybe the problem is when doing
> inline optimization?
It could be, but I don't think inlines are done with -O2 (have to
double-check the egcs docs for this one....nope...no inlines with -O2 (-O3
includes inlining functions). Also, loop unrolling isn't done at the -O2
level, not that it would be causing this, but it was fun to say it :-)
I may try building the next egcs release without the Haifa scheduler again
to see if that's still buggy (wouldn't doubt it). I haven't run parallel
tests like I usually do (Haifa vs. non- on the same source),
unfortunately, so I don't know yet if this is a compiler-wide problem or
just related to the "improved" scheduler.
C
Reply to: