paul@cacilj.UUCP (Paul Close) (02/09/89)
I am confirming what has already been posted here: the cc1 program in
stage2 blows up on several files with signals 6 or 11. I used optimization
only with gcc; I used no flags with the plain cc. The object files that
*were* produced compared perfectly.
There are 14 failures: 6 SIGABRT (abort), 8 SIGSEGV (segmentation violations).
here's a transcript after doing a second `make -k CC=gcc CFLAGS="-O -Bstage2/"'
(to only show the ones that failed):
---------------------------------------------------------
stage2/gcc -O -Bstage2/ -sun4 -c toplev.c
stage2/gcc: Program cc1 got fatal signal 6.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c tree.c
stage2/gcc: Program cc1 got fatal signal 11.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c stor-layout.c
stage2/gcc: Program cc1 got fatal signal 6.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c fold-const.c
stage2/gcc: Program cc1 got fatal signal 11.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c expr.c
stage2/gcc: Program cc1 got fatal signal 6.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c stmt.c
stage2/gcc: Program cc1 got fatal signal 11.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c expmed.c
stage2/gcc: Program cc1 got fatal signal 6.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c explow.c
stage2/gcc: Program cc1 got fatal signal 11.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c varasm.c
stage2/gcc: Program cc1 got fatal signal 11.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c cse.c
stage2/gcc: Program cc1 got fatal signal 11.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c loop.c
stage2/gcc: Program cc1 got fatal signal 11.
*** Error code 1
stage2/gcc -O -Bstage2/ -DGCC_INCLUDE_DIR=\"/usr/local/lib/gcc-include\" \
-DGPLUSPLUS_INCLUDE_DIR=\"/usr/local/lib/g++-include\" -c cccp.c
stage2/gcc: Program cc1 got fatal signal 6.
*** Error code 1
stage2/gcc -O -Bstage2/ -sun4 -c cexp.c
stage2/gcc: Program cc1 got fatal signal 6.
*** Error code 1
make: Warning: Target `all' not remade because of errors
---------------------------------------------------------
Here's an adb listing from the last compile, cexp.c:
---------------------------------------------------------
core file = core -- program ``cc1''
SIGIOT 6: abort
$c
_kill() + 8
__exit_handlers(?)
_emit_move_insn(0xdcbb8,0xa2bb8,0xd41f8,0xdd278,0xdcbb0,0xdcbb8) + a
_convert_move(0xdcbb8,0xa2bb8,0x0,0x0,0xffffffff,0x8) + 514
_emit_library_call(0x2,0x3,0x2,0xf7ffdcb8,0x3,0xdcbb0) + d0
_expand_binop(0xdcba0,0xdcb98,0xdcb88,0x3,0x0,0xdc5a8) + 460
_expand_divmod(0xdcb88,0x42,0xffffffff,0xdcb88,0xdcb98,0x0) + 6f4
_expand_expr(0x0,0x42,0x0,0x0,0xa6c18,0xa2908) + 1708
_store_field(0xd5d90,0x20,0x0,0x3,0xa6c18,0x0) + 58
_expand_assignment(0x0,0x20,0x0,0x4,0xda9e0,0xda9f8) + 218
_expand_expr(0x1,0x37,0x0,0x0,0xa2bb8,0xa2908) + 21e4
_expand_expr_stmt(0xa6c38,0xb8,0x16a,0x1be,0x0,0x80) + b4
_yyparse(0x38,0xf7fff370,0x0,0x2,0xb17e0,0xf) + 19d8
_compile_file(0xf7fffa7b,0x1be90,0x1000,0x3,0xf7fffaab,0x0) + 98c
_main(0x0,0xf7fffa7b,0x8,0x1c038,0x0,0x0) + 878
---------------------------------------------------------
Hope this helps. I can provide more info on request....
--
Paul Close paul@cacilj.CTS.COM ...!{uunet, ucsd, crash}!cacilj!paul
The Obi-wan Kenobi method: "Use the Source, Luke" -Jim Fultonpaul@cacilj.UUCP (Paul Close) (02/14/89)
In article <893@cacilj.UUCP> I write: | I am confirming what has already been posted here: the cc1 program in | stage2 blows up on several files with signals 6 or 11. I used optimization | only with gcc; I used no flags with the plain cc. The object files that | *were* produced compared perfectly. | There are 14 failures: 6 SIGABRT (abort), 8 SIGSEGV (segmentation violations). I have found that this problem occurs when I compile using CFLAGS="-O -Bstage(n)/", but NOT when CFLAGS="-g -O -Bstage(n)/". Definitely sounds fishy.... Seems like somehow the -g flag modifies the -O flag's behaviour. -- Paul Close paul@cacilj.CTS.COM ...!{uunet, ucsd, crash}!cacilj!paul The Obi-wan Kenobi method: "Use the Source, Luke" -Jim Fulton