The current 3.3 branch configured for any mips-sgi-irix6o32 target fails to bootstrap even with this patch http://gcc.gnu.org/ml/gcc-patches/2003-05/msg02284.html applied: /vol/gcc/obj/gcc-3.3.1-20030527/6.5o32-cc-as/gcc/xgcc -B/vol/gcc/obj/gcc-3.3.1-20030527/6.5o32-cc-as/gcc/ -B/vol/gcc/share/mips-sgi-irix6.5o32/bin/ -B/vol/gcc/share/mips-sgi-irix6.5o32/lib/ -isystem /vol/gcc/share/mips-sgi-irix6.5o32/include -c -DSkip_f2c_Undefs -DAllow_TYQUAD -I. -I/vol/gnu/src/gcc/gcc-3.3-branch-dist/libf2c/libI77 -I.. -I/vol/gnu/src/gcc/gcc-3.3-branch-dist/libf2c/libI77/.. -DHAVE_CONFIG_H -W -Wall -O2 -g /vol/gnu/src/gcc/gcc-3.3-branch-dist/libf2c/libI77/uio.c -DPIC cc1: warning: `-g': unknown or unsupported -g option In file included from /vol/gcc/obj/gcc-3.3.1-20030527/6.5o32-cc-as/gcc/include/stdio.h:18, from /vol/gnu/src/gcc/gcc-3.3-branch-dist/libf2c/libI77/fio.h:2, from /vol/gnu/src/gcc/gcc-3.3-branch-dist/libf2c/libI77/uio.c:3: /usr/include/internal/stdio_core.h: In function `snprintf': /usr/include/internal/stdio_core.h:307: error: Attempt to delete prologue/epilogue insn: (insn 41 35 42 0 0 (set (mem:SI (plus:SI (reg/f:SI 29 sp) (const_int 40 [0x28])) [0 S4 A32]) (reg:SI 6 a2)) -1 (nil) (nil)) /usr/include/internal/stdio_core.h:307: internal compiler error: in propagate_one_insn, at flow.c:1639 The same problem exists when using gas 2.13.91 (and configuring --with-gnu-as/--with-as). This is a regression from 3.2. Environment: System: IRIX lyra 6.5 04100802 IP22 host: mips-sgi-irix6.5o32 build: mips-sgi-irix6.5o32 target: mips-sgi-irix6.5o32 configured with: /vol/gnu/src/gcc/gcc-3.3-branch-dist/configure --prefix=/vol/gcc --with-local-prefix=/vol/gcc --disable-nls mips-sgi-irix6.5o32 How-To-Repeat: Bootstrap the compiler as above.
Confirming my own report. Assigned to Richard Sandiford who mentioned he had a least a workaround for this issue: http://gcc.gnu.org/ml/gcc-patches/2003-05/msg02236.html Richard, do you have a preliminary patch so I can continue working on other mips-sgi-irix6o32 issues?
Created attachment 4100 [details] GP optimisation patch that happens to work around the problem Hmm, I notice this bug report is against 3.3. I'm afraid the patch won't apply there and it'd be far too invasive anyway. Still, I've attached it here in case it comes in useful. NB, I posted this patch for approval yesterday but it doesn't seem to have shown up in the archives.
Subject: Re: ICE (Attempt to delete prologue/epilogue insn) during mips-sgi-irix6o32 bootstrap > Hmm, I notice this bug report is against 3.3. I'm afraid the patch won't > apply there and it'd be far too invasive anyway. Still, I've attached it > here in case it comes in useful. Thanks. While bootstrap on the 3.3 branch breaks in libf2c, mainline bootstrap already breaks in stage2. I'll give it a try. > NB, I posted this patch for approval yesterday but it doesn't seem > to have shown up in the archives. Indeed: there's a considerable gap in the archive. I got that patch via the list, though. Rainer
Added Eric to Cc: per his request.
This patch (Richard Sandiford, Various gp-related mips optimisations, of 27 May 2003, unfortunately still missing from the web archive) allows me to successfully bootstrap mainline for various mips-sgi-irix6o32 configurations. Any chance to get this checked in? And what about the 3.3 branch? Rainer
Subject: Re: ICE (Attempt to delete prologue/epilogue insn) during mips-sgi-irix6o32 bootstrap > This patch (Richard Sandiford, Various gp-related mips optimisations, of > 27 May 2003, unfortunately still missing from the web archive) allows me to > successfully bootstrap mainline for various mips-sgi-irix6o32 configurations. > > Any chance to get this checked in? And what about the 3.3 branch? I approved the patch for mainline. Not a chance for 3.3. It's too much. -eric
Hello, was this bug a 3.3 regression? Otherwise, I presume we can close this as "fixed in 3.4." Thanks, Dara
Subject: Re: ICE (Attempt to delete prologue/epilogue insn) during mips-sgi-irix6o32 bootstrap dhazeghi@yahoo.com writes: > was this bug a 3.3 regression? Otherwise, I presume we can close this as "fixed in 3.4." Thanks, The mips-sgi-irix6*o32 configurations were only introduced in 3.3 (and used to work for some time before the release, but got broken lately), so this doesn't count as a 3.3 regression ;-( Rainer
Created attachment 4269 [details] Proposed patch for this PR and for 11084 Rainer, could you try the attached patch? I'm testing it on mipsel-linux-gnu as a potential fix for 11084. Since that PR is a regression, the patch might be accepted for 3.3.1. Richard
http://gcc.gnu.org/ml/gcc-patches/2003-06/msg02766.html
Subject: Re: ICE (Attempt to delete prologue/epilogue insn) during mips-sgi-irix6o32 bootstrap > Rainer, could you try the attached patch? I'm testing it on > mipsel-linux-gnu as a potential fix for 11084. Since that PR > is a regression, the patch might be accepted for 3.3.1. I have tried bootstrapping mips-sgi-irix6.5o32 on the 3.3 branch before this patch, and curiously it worked even without (both with native as and gas 2.14). I'll try another bootstrap from the branch when checking some of my O32 patches from mainline on the 3.3 branch. Thanks. Rainer