To: pgcc AT delorie DOT com
Subject: [1.1.2] A non-Haifa bootstrap failure
X-Url: http://www.cs.ucsb.edu/~jerry/
X-Tom-Swiftie: "I don't want any champagne!" Tom said, blowing his top.
Mime-Version: 1.0 (generated by tm-edit 7.108)
Content-Type: text/plain; charset=US-ASCII
From: jerry AT cs DOT ucsb DOT edu (Jerry James)
Date: 12 Apr 1999 15:53:34 -0700
Message-ID: <tpremlpjvrl.fsf@cs.ucsb.edu>
Lines: 62
X-Mailer: Gnus v5.6.45/XEmacs 20.4 - "Emerald"
X-Face:  +5(Pfr,;N<RBl'|Xv5zrC\p$;u*nfOiN-E:PeZJda+Y!RC[QkTvF*GR',H5EV,-b;n*%1Hb
 {ZR|hIzas}[;QCvbj=M*ht_-Yp72~mNX0ZVO\$t-6X:mauuY-YrLYDZ$,SnTL^ni{aFj;i1FgDEo,'
 p;B!2B!_F=,`R%TDZxPy@+{&qNKT:9-xg(Z**>>q#6NT,Qi5^TQh-MaUnz#kGN~OW[CQj~RS+sIor(
 '_8K^f9u^Y#.N`>9oKN$\JpI
Reply-To: pgcc AT delorie DOT com


I am not on this mailing list, although I do browse the web archive once 
a week or so.  I saw that a number of people trying to compile pgcc
1.1.2 with the Haifa scheduler were getting ICEs.  Well, here's one for
a non-Haifa bootstrap.

Compiling pgcc 1.1.2 with:
  BOOT_CFLAGS = "-O6 -march=pentium -mcpu=pentium"
  CFLAGS = "-O6 -march=pentium -mcpu=pentium" \
  LIBCFLAGS = "-g -O6 -march=pentium -mcpu=pentium" \
  LIBCXXFLAGS = "-g -O6 -march=pentium -mcpu=pentium -fno-implicit-templates"

I am bootstrapping with egcs 1.1.2 (seems reasonable, no?).  The first
stage compiler builds fine.  However, the bootstrap dies while trying to 
build the very first file for stage 2.  I get this:

------------------------------------------------------------------------
stage1/xgcc -Bstage1/  -DIN_GCC    -O6 -march=pentium -mcpu=pentium  -DHAVE_CONFIG_H      -I. -I../../egcs-1.1.2/gcc -I../../egcs-1.1.2/gcc/config -c ../../egcs-1.1.2/gcc/cexp.c
../../egcs-1.1.2/gcc/sched.c:5296: Internal compiler error in function update_flow_info
------------------------------------------------------------------------

I tried compiling that file by hand, to see if some particular
optimization was causing the problem.  I get the same results with -O2
through -O6.  However, with just -O, I get the following:

------------------------------------------------------------------------
stage1/xgcc -Bstage1/ -DIN_GCC -O -march=pentium -mcpu=pentium -DHAVE_CONFIG_H -I. -I../../egcs-1.1.2/gcc -I../../egcs-1.1.2/gcc/config -c ../../egcs-1.1.2/gcc/cexp.c
/usr/share/bison.simple: In function `yyparse':
/usr/share/bison.simple:692: Could not split insn
(insn 517 515 524 (parallel[ 
            (set (mem/s:DI (plus:SI (reg:SI 6 %ebp)
                        (const_int -2812)))
                (sign_extend:DI (reg:SI 2 %ecx)))
            (clobber (reg:SI 0 %eax))
        ] ) 110 {extendsidi2} (insn_list 514 (nil))
    (expr_list:REG_DEAD (reg:SI 2 %ecx)
        (nil)))
../../egcs-1.1.2/gcc/toplev.c:1438: Internal compiler error in function fatal_insn
------------------------------------------------------------------------

and with no optimization, I get similar output:

------------------------------------------------------------------------
stage1/xgcc -Bstage1/ -DIN_GCC -march=pentium -mcpu=pentium -DHAVE_CONFIG_H -I. -I../../egcs-1.1.2/gcc -I../../egcs-1.1.2/gcc/config -c ../../egcs-1.1.2/gcc/cexp.c
/usr/share/bison.simple: In function `yyparse':
/usr/share/bison.simple:692: Could not split insn
(insn 499 498 2342 (parallel[ 
            (set (mem:DI (plus:SI (reg:SI 6 %ebp)
                        (const_int -2876)))
                (sign_extend:DI (reg:SI 2 %ecx)))
            (clobber (reg:SI 0 %eax))
        ] ) 110 {extendsidi2} (nil)
    (nil))
../../egcs-1.1.2/gcc/toplev.c:1438: Internal compiler error in function fatal_insn
------------------------------------------------------------------------

I'll try bootstrapping again, but build the first stage compiler with
only -O2, in case this is an egcs 1.1.2 bug with -O6 (really -O3).  I'll 
let you know if it makes any difference.
-- 
Jerry James
Email: jerry AT cs DOT ucsb DOT edu
WWW:   http://www.cs.ucsb.edu/~jerry/