Hello,
I've looked into my gcc-3.2 problem again and found out that gcc-3.2
with -march=athlon produces the problem I described in every detail in a
previos mail (wirebox example).
What made this so tedious to find (the problem was appearing and
disappearing arbitrarily) is an inconsistency in the Makefiles. When I
run a global make from xc/xc it uses the optimization options I
specified in host.def (-O2 -march=athlon). When I run make locally in a
subdirectory (I tried lib/GL/mesa/src/drv[/radeon]) it uses only -O2.
I looked into the local Makefile and found that the definition of CFLAGS
appears twice in the Makefile. The first one specifies -O2 -march=athlon
in CDEBUGFLAGS, the second one specifies only -O2. So in effect a local
make uses only -O2. Doing a global make CDEBUGFLAGS is specified on make
the command line and of make for all subdirectories in xc/xc/xmakefile:
for i in $(SUBDIRS) ;\
do \
echo "making" all "in $(CURRENT_DIR)/$$i..."; \
$(MAKE) -C $$i $(MFLAGS) $(PARALLELMFLAGS) CDEBUGFLAGS="$(CDEBUGFLAGS)"
all; \
done
This overrides the variable definitions of CDEBUGFLAGS in all
subdirectories.
Now we still have to find the exact cause of the problem with
-march=athlon. First of all, can anyone reproduce it?
Best regards,
Felix
__\|/__ ___ ___ ___
__Tsch��_______\_6 6_/___/__ \___/__ \___/___\___You can do anything,___
_____Felix_______\�/\ \_____\ \_____\ \______U___just not everything____
[EMAIL PROTECTED] >o<__/ \___/ \___/ at the same time!
-------------------------------------------------------
This sf.net email is sponsored by:ThinkGeek
Welcome to geek heaven.
http://thinkgeek.com/sf
_______________________________________________
Dri-devel mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/dri-devel
