internal compiler error: Segmentation fault

["Lee Rhodes" <lee at rhoba dot com>]

Hello,
  I am attempting to build gcc 4.3.0-20070622 on i686-pc-cygwin 1.5.24-2.
Cygwin has been updated with the cygwin-inst-20070616.tar.bz2 in order to
obtain Dave Korn's stdio.h / newlib fix (which I have verified is there).  

  The config options are:
  --prefix=/usr/local/gcc-4.3-20070622 
  --disable-shared 
  --enable-sjlj-exceptions 
  --enable-bootstrap 
  --enable-libgomp 
  --enable-libmudflap 
  --enable-threads=posix 
  --enable-languages='c c++' 

	=== gcc Summary ===

  # of expected passes        46896
  # of unexpected failures    67
  # of unexpected successes   2
  # of expected failures      134
  # of untested testcases     42
  # of unsupported tests      447

However, looking more closely at the unexpected fails, I noticed that there
were 12 more than Tim Prince's idential build/test. These fails all have the
following signature in the gcc.log (I have deleted the paths and duplicate
lines to make it easier to read):

  Executing on host: xgcc -O2 -w -fno-show-column -o 20001226-1.o
gcc.c-torture/compile/20001226-1.c
    gcc.c-torture/compile/20001226-1.c: In function 'foo': 
    gcc.c-torture/compile/20001226-1.c:23: internal compiler error:
Segmentation fault
  FAIL: gcc.c-torture/compile/20001226-1.c (internal compiler error)

I can duplicate these failures more precisely with the following 
  make check-gcc RUNTESTFLAGS="compile.exp=20001226-1.c"

It is perhaps notable that this particular torture test does PASS at
optimization settings of -O0 and -O1, but always has the same failure
signature at Optimizations of -O2, -O3 and -Os.  Also I noted that the only
one of these tests that failed on Tim's build was the -Os, but it was a
different failure and not the Segmentation fault I am getting. 

The only other unusual messages appear at the end of the build standard
error output:

  configure: error: none of the known symbol names works
  make[1]: *** [configure-target-libmudflap] Error 1
  make: *** [bootstrap] Error 2
    9 [unknown (0xCA4)] conftest 7216 _cygtls::handle_exceptions: Error
while dumping state (probably corrupted stack)

I am stuck.  I don't know what a Segmentation fault is, but since Tim got
his build to test without any segmentation faults, I'm not sure it is a bug.
I suspect that perhaps my Cygwin may be corrupted or missing something.  

Does anyone have any suggestions?

Thanks,

Lee.