Re: New trunk breakage on mips-sgi-irix6.2 stage1 compiling libgcc2

["Kaveh R. Ghazi" <ghazi at caip dot rutgers dot edu>]

 > From: Alexandre Oliva <aoliva@redhat.com>
 > 
 > On May  3, 2001, "Kaveh R. Ghazi" <ghazi@caip.rutgers.edu> wrote:
 > 
 > >> 2001-05-03  Alexandre Oliva  <aoliva@redhat.com>
 > >> 
 > >> Re-installed 2001-01-09's patch:
 > >> * hwint.h (HOST_BITS_PER_WIDE_INT, HOST_WIDE_INT): Use long long
 > >> if it's wider than long and the target's long is wider than the
 > >> host's.
 > 
 > The culprit is definitely this one (actually, it just exposes
 > pre-existing bugs, but who cares? :-)
 > 
 > Unfortunately, I don't have access to an IRIX box at the moment, and
 > I'm about to take off for tonight, so I won't be able to do much about
 > it in the next 24 hours :-(

Ok, though lack of access to IRIX shouldn't stop you since I was able
to trigger it in a cross.



 > If you have a bit of spare time and would like to help, the trick to
 > solve this problem is generally to go after GEN_INT expressions in the
 > .md, .c and .h files and see if they could possibly be generating RTL
 > CONST_INTs that are not valid sign-extensions within the mode of the
 > operation in which they're used, and introduce a trunc_int_for_mode()
 > to make sure it's a sign-extension.  Don't worry about peepholes or
 > values used only for output; expands and splits are the places in
 > which you'll find problems.

I'll poke at it, but this isn't my area of expertise. :-)


 > 
 > I've just posted two patches that fix problems related with
 > sign-extension of CONST_INTs in the middle end; perhaps they might
 > help mips too?

I tried them and it didn't help. :-(

--
Kaveh R. Ghazi			Engagement Manager / Project Services
ghazi@caip.rutgers.edu		Qwest Internet Solutions