This is the mail archive of the
mailing list for the GCC project.
Re: PING: gimplifier ICE fix
You're not always being very constructive yourself, you know...
Anyway, here's something you could do.
- get preprocessed sources of a full gcc build tree. You can do
that by bootstrapping with -save-temps, and then collect all
the .i and .ii files in one place.
- do "for $f in *.i; \
do $build/gcc/cc1 -quiet -fdump-tree-optimized -O2 $f; \
done". Do the same for all .ii files. Do this for a patched
and an unpatched compiler, and compare all the .optimized files.
If they are identical, you've taken the first hurdle.
- hack the test suite scripts to dump the .optimized tree dump in
some directory (but *not* the $build/gcc/testsuite/ directory)
Do this again for the patched and unpatched compilers, and again
Sounds like a plan, Steven! Go to it. I don't have the time.
See, that's what I mean. I'm trying to help you somewhat, and all
I get is this kind of arrogant answer. Go fuck yourself.
Wow, so much love! I think I'll stop now and absorb it all and let
you guys sort out your gimplifier woes...
There are no gimplifier woes though. That's the point. It's only because
objc is a bit "loose" with it's types, and then gets angry when other
parts of the compiler rely on it, that we have this problem.