Re: MIPS ICE with CONST_HIGH_PART sign-extension problem

[Jim Wilson <wilson at codesourcery dot com>]

On Mon, 2009-04-27 at 19:49 +0000, Joseph S. Myers wrote:
> Since this test isn't at all MIPS-specific, it should go in 
> gcc.c-torture/compile rather than gcc.target/mips, so it can potentially 
> show up bugs on other targets or with other optimization options.

The testcase won't work for all targets, and I'm not sure how to handle
that.  This isn't a problem in the gcc.target/mips directory.  If the
testcase gets modified to work for other targets, then it probably won't
still trigger the MIPS backend bug, so that isn't a good solution
either.

For instance, trying the test with avr-elf, I get
khazaddum$ ./xgcc -B./ -O -S tmp.c
tmp.c: In function âsub2â:
tmp.c:18: warning: overflow in implicit constant conversion

I suppose I could add a test for ilp32, but I'm not sure if that is
sufficient, and it will take some time to test.

Jim