Trunk broken for MMIX, CRIS, i960, m32r, (v850, ARM)

[Hans-Peter Nilsson <hp at bitrange dot com>]

Using *-unknown-elf, except mmix-knuth-mmixware.

Worked LAST_UPDATED Thu Jul  3 10:03:12 GMT 2003
Broken LAST_UPDATED Fri Jul  4 11:17:49 GMT 2003
*Still* broken, at least on MMIX and CRIS
    at LAST_UPDATED Fri Jul  4 22:17:01 GMT 2003
(i.e. including Jeff's and Zdenek's patches, which looked
promising)

The targets in parentheses didn't build last time either (at
"Worked"), but they all see something like this (for
mmix-knuth-mmixware)  compiling different files:

...
x/gcc/libgcc2.c: In function `__gcc_bcmp':
x/gcc/libgcc2.c:1519: error: head insn 111 for block 0 not found in the insn stream
x/gcc/libgcc2.c:1519: internal compiler error: Segmentation fault

I see related messages for sparc64, and the regression tester
reported same breakage for powerpc-apple-darwin6.3.  A wild
guess says it'd also be visible with powerpc-unknown-eabisim.

<rant>
Hey, what's the big hurry with patches that people don't have
the time to run test them with a couple of *complete
cross-toolchains including simulators* available at their
fingertips at no extra charge:
<URL:http://gcc.gnu.org/simtest-howto.html>.  Nag, nag, nobody
cares; "*these* patches are good enough that they should just
need a bootstrap+check on my PC, and only for face value".
</rant>

Please help improve the simtest instructions, if that's the
fault.

Hmm, I'll see to adding instructions for MMIX, despite that
you'd have to go get a separate package for the simulator
(Knuth's "MMIXware", see
<URL:http://www-cs-faculty.stanford.edu/~knuth/mmix-news.html>).
A separate section seems in order for such targets requiring
external components.

brgds, H-P