../../../libio/stream.cc:60: Internal error: Segmentation fault.

[John David Anglin]

With todays and yesterdays CVS source, I get the following error and under
i686 linux and hpux 10.20:

/home/dave/gnu/gcc-2.97/objdir/gcc/g++ -B/home/dave/gnu/gcc-2.97/objdir/gcc/ -nostdinc++ -isystem /home/dave/gnu/gcc-2.97/libstdc++ -isystem /home/dave/gnu/gcc-2.97/libstdc++/std -isystem /home/dave/gnu/gcc-2.97/libstdc++/stl -isystem /home/dave/gnu/gcc-2.97/libio -isystem /home/dave/gnu/gcc-2.97/objdir/i686-pc-linux-gnu/libio -L/home/dave/gnu/gcc-2.97/objdir/i686-pc-linux-gnu/libstdc++ -B/usr/local/i686-pc-linux-gnu/bin/ -B/usr/local/i686-pc-linux-gnu/lib/ -isystem /usr/local/i686-pc-linux-gnu/include -c -g -O3 -fvtable-thunks -D_GNU_SOURCE -fno-implicit-templates -I. -I../../../libio -nostdinc++ -D_IO_MTSAFE_IO ../../../libio/stream.cc
../../../libio/stream.cc: In function `char* dec(long int, int)':
../../../libio/stream.cc:60: Internal error: Segmentation fault.
Please submit a full bug report.

Configure was as follows:

CFLAGS="-g -O3"; export CFLAGS
BOOT_CFLAGS="-g -O3"; export BOOT_CFLAGS
../configure --with-gnu-as --with-gnu-ld --enable-threads=posix --disable-libstdcxx-v3 && make bootstrap-lean && make -k RUNTESTFLAGS="--target_board 'unix{-pth

I first noticed the error testing new patches to improve and simplify
the detection of base register operands on the pa port.  On the pa, the
fault occurs here at line 10066 in expr.c:

do_compare_and_jump+604:        LDWX,S          r19(r20),r20

The register r19 and r20 are 0x7ae97400 and 0x46954, respectively.  Thus,
the base and index registers are reversed.  This could be a problem with
the patches which I am testing.  However, the same fault occurs with an
unpatched i686 linux build:

Program received signal SIGSEGV, Segmentation fault.
0x81df1aa in do_compare_and_jump (exp=0x40536de0, signed_code=GT,
    unsigned_code=GTU, if_false_label=0x40533e80, if_true_label=0x0)
    at ../../gcc/expr.c:10066
10066                 > GET_MODE_BITSIZE (TREE_TYPE (TREE_OPERAND (exp, 1))))))
(gdb) disass 0x81df1aa 0x81df1b0
Dump of assembler code from 0x81df1aa to 0x81df1b0:
0x81df1aa <do_compare_and_jump+842>:    mov    0x843d520(,%eax,4),%eax
End of assembler dump.
(gdb) info reg eax
eax            0x40155400       1075139584

I think one of the changes in the last week has caused this.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6605)