Re: what's up with gts?
On Fri, Jun 20, 2003 at 10:01:19AM +0100, Philip Blundell wrote:
> This kind of thing usually means an ICE. You need to run a build by
> hand on debussy, and remove libtool's ultra-helpful ">/dev/null 2>&1"
> from the commandline so that we can actually see the error messages.
Ok. Did that. The result is exactly what I was expecting:
(sid)mmagallo@debussy:~/gts-0.7.1/src$ gcc -DHAVE_CONFIG_H -I. -I. -I.. -I.. -I/usr/include -DG_LOG_DOMAIN=\"Gts\" -O2 -Wall -Wall -Werror-implicit-function-declaration -Wstrict-prototypes -Wmissing-prototypes -Wmissing-declarations -I/usr/include/glib-1.2 -I/usr/lib/glib/include -c predicates.c -MT predicates.lo -MD -MP -MF .deps/predicates.TPlo -o predicates.o
predicates.c: In function `insphereexact':
predicates.c:2368: internal compiler error: in change_address_1, at emit-rtl.c:2017
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.
The preprocessed source can be found in
The function in question is just a largish inlined (via macros)
computation. My guess is the compiler is register-starved and can't
find its way out of that mess.
Can I help further?