Re: Regressions from Re: [PATCH] remove the unsigned_type language hook

[Hans-Peter Nilsson <hans-peter dot nilsson at axis dot com>]

> Date: Tue, 15 May 2007 10:09:23 +0100
> From: "Rafael Espindola" <espindola@google.com>

> > I guess something was flawed in your regression testing,
> > comparing old logs or something.
> 
> I have just bootstraped and tested gcc trunk r124715 and I don't see
> any of these:
> 
> > Running /tmp/hpautotest-gcc1/gcc/gcc/testsuite/gcc.dg/dg.exp ...
> > FAIL: gcc.dg/Wno-pointer-sign.c (test for excess errors)
> > FAIL: gcc.dg/Wpointer-sign-Wall-no.c (test for excess errors)
> > FAIL: gcc.dg/Wpointer-sign-Wall.c  (test for warnings, line 8)
> > FAIL: gcc.dg/Wpointer-sign-Wall.c (test for excess errors)
> > FAIL: gcc.dg/Wpointer-sign-pedantic-no.c (test for excess errors)
> > FAIL: gcc.dg/Wpointer-sign-pedantic.c  (test for warnings, line 8)
> ....
> 
> What architecture are you running the tests on?

As I said: "this change causes the following regressions for
cris-elf (only suspect in the range 124713:124716) and I'd be
surprised to see all pass for any target".  The cris-elf is my
usual cross-test target, see instructions in simtest-howto.html
(though dejagnu-1.4.4 lacks the baseboard, instructions in
<http://gcc.gnu.org/ml/gcc-patches/2007-02/msg01571.html>).

If you don't like that target, pick another one.  Hm, perhaps
you won't see this on a target where char is default unsigned?
What target did *you* test on?  I don't see you mentioned the
target, but perhaps I missed it just like you did.

brgds, H-P