Re: PATCH: mips16 function attributes, version N+1

[Sandra Loosemore <sandra at codesourcery dot com>]

David Daney wrote:
> Sandra Loosemore wrote:
> > Richard Sandiford wrote:
> >
> > > Unfortunately, I think removing the code _will_ need a retest.  OK with
> > > that change on top of the other ones.
> >
> > OK, fixed up as suggested, retested, and committed now.  For the 
> > record, I'm attaching the final version of the patch as committed.
> >
> > -Sandra
>
> How was this tested?

For testing the MIPS part, I was using a mipsisa32r2-elfoabi 
configuration, built on i686-pc-linux-gnu.  I ran the gcc and g++ 
testsuites with the following combinations of options:

  default
  -mfp64
  -msoft-float
  -EL
  -EL -mfp64
  -EL -msoft-float
  -mips16
  -mips16 -mfp64
  -mips16 -msoft-float
  -mips16 -EL
  -mips16 -EL -mfp64
  -mips16 -EL -msoft-float

For testing -mflip-mips16, I was using only the gcc testsuite, and ran 
it with all the same combinations of other flags listed above.  I 
investigated every test case that was failing with -mflip-mips16 that 
wasn't already failing with one of the {-mips16, -mnomips16} base 
configurations.

I also did a full bootstrap and regression test on i686 and an ARM build 
and test, to verify that the non-MIPS-specific support pieces were doing 
sensible things on other targets.

> It causes bootstrap failure due to an ICE building libstdc++ on 
> mipsel-linux.
>
> See: http://gcc.gnu.org/ml/gcc/2007-09/msg00143.html

Hrrmmm.  I'm not sure I can duplicate this build environment.  Any 
chance you could get me a backtrace from the point of the ICE so I could 
take a quick look at where the error is happening?  Otherwise I suppose 
I'll have to revert at least the MIPS-specific part of the patch while 
we investigate.  :-(

-Sandra