egcs-CVS19981102 powerpc-ibm-aix4.1.4.0 bootstrap failure

[Franz Sirl]

Hi,

Try the patch posted by Herman ten Brugge in egcs-patches "fix error in
emit-rtl.c". Despite Jeff's comment it seems to have an effect and fixes
the testsuite regression I reported on powerpc-unknown-linux-gnu

Franz.


At 16:25 03.11.98 , Kaveh R. Ghazi wrote:
>Snapshot:		egcs-CVS19981102
>Platform:		powerpc-ibm-aix4.1.4.0
>Stage1 compiler:	cc
>Assembler:		native
>Linker:			native
>BOOT_CFLAGS:		-g -O2 -mminimal-toc -pipe
>
>
>	The build dies after the 3 stage bootstrap when compiling libio:
>
> > /caip/u47/ghazi/gcc-testing/aix-test/egcs-CVS19981102/gcc/xgcc
> > 	-B/caip/u47/ghazi/gcc-testing/aix-test/egcs-CVS19981102/gcc/
> > 	-B/usr/local/powerpc-ibm-aix4.1.4.0/bin/ -c -g -msoft-float -I. -I.
> > 	floatconv.c -o pic/floatconv.o
> > /caip/u47/ghazi/gcc-testing/aix-test/egcs-CVS19981102/gcc/xgcc
> > 	-B/caip/u47/ghazi/gcc-testing/aix-test/egcs-CVS19981102/gcc/
> > 	-B/usr/local/powerpc-ibm-aix4.1.4.0/bin/ -c -g -msoft-float -I. -I.
> > 	floatconv.c
> > final.c:2915: Internal compiler error in function final_scan_insn
> > make[4]: *** [floatconv.o] Error 1
>
>
>	Here's the source code where it dies:
>
> >         /* If the template is the string "#", it means that this insn must
> >            be split.  */
> >         if (template[0] == '#' && template[1] == '\0')
> >           {
> >             rtx new = try_split (body, insn, 0);
> >  
> >             /* If we didn't split the insn, go away.  */
> >             if (new == insn && PATTERN (new) == body)
> >               fatal_insn ("Could not split insn", insn);
> >  
> > #ifdef HAVE_ATTR_length
> >             /* This instruction should have been split in shorten_branches,
> >                to ensure that we would have valid length info for the
> >                splitees.  */
> >             abort ();     <<<<<---- line 2915 ----
> > #endif
> >  
> >             new_block = 0;
> >             return new;
> >           }
> > 
>
>	If there's any other useful info I can provide, let me know.
>And if someone else reproduces this, I'd like to hear confirmation.
>
>		Thanks,
>		--Kaveh
>--
>Kaveh R. Ghazi			Engagement Manager / Project Services
>ghazi@caip.rutgers.edu		Icon CMT Corp.