more hpux fun (ICE building libgcc2.c in __fixunsdfdi)

[Joe Buck]

Still can't bootstrap hppa2.0-hp-hpux10.20.

We get further than before (the duplicate-include problem is gone),
but still bomb early in stage 2.  The latest failure is

./xgcc -B./ -B/u/jbuck/cvs-gcc-hp/hppa2.0-hp-hpux10.20/bin/ -isystem \
/u/jbuck/cvs-gcc-hp/hppa2.0-hp-hpux10.20/include -O2   -DIN_GCC \
-W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes \
-isystem ./include  -fPIC -g1  -DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED  \
-I. -I. -I/u/jbuck/gnu/src/cvs/gcc/gcc -I/u/jbuck/gnu/src/cvs/gcc/gcc/. \
-I/u/jbuck/gnu/src/cvs/gcc/gcc/config \
-I/u/jbuck/gnu/src/cvs/gcc/gcc/../include -DUSE_COLLECT2 -DL_fixunsdfdi \
-c /u/jbuck/gnu/src/cvs/gcc/gcc/libgcc2.c -o libgcc/./_fixunsdfdi.o
/u/jbuck/gnu/src/cvs/gcc/gcc/libgcc2.c: In function `__fixunsdfdi':
/u/jbuck/gnu/src/cvs/gcc/gcc/libgcc2.c:949: Internal compiler error in propagate
_one_insn, at flow.c:3790

The relevant comment says

      /* If we're trying to delete a prologue or epilogue instruction
	 that isn't flagged as possibly being dead, something is wrong.
	 But if we are keeping the stack pointer depressed, we might well
	 be deleting insns that are used to compute the amount to update
	 it by, so they are fine.  */

Clues, Jeff?