[Bug middle-end/25125] [4.1 Regression] (short) ((int)(unsigned short) + (int)) is done in the wrong type

[dave at hiauly1 dot hia dot nrc dot ca]

------- Comment #20 from dave at hiauly1 dot hia dot nrc dot ca  2005-12-27 19:27 -------
Subject: Re:  [4.1 Regression] (short) ((int)(unsigned short) + (int)) is done
in the wrong type

> > I am having tough time building the compiler for hppa2.0w-hp-hpux11.11.
> > These days gcc won't even allow you to build cc1 without an assembler and such.
> 
> That's a regression.  It always used to be possible to build
> cc1 on x86, etc.  I'll take a look.

This seems to work for me on x86 linux:

../gcc/configure --with-gnu-as --disable-shared --disable-nls
--prefix=/home/dave/opt/gnu  --target=hppa2.0w-hp-hpux11.11 &&
make &&

However, I'm getting different tree code with the x86 cross.

;; Function f (f)

f (a)
{
  int D.1491;

<bb 2>:
  if ((short int) a > 0) goto <L4>; else goto <L1>;

<L4>:;
  D.1491 = 0;
  goto <bb 4> (<L2>);

<L1>:;
  D.1491 = (int) (short unsigned int) (short int) (a + 32768);

<L2>:;
  return D.1491;

}

Not sure why?  The native hppa build generates the same code using
the stage1 compiler.  I have a few patches in my tree.  The only one
that might be relevant is the fix for PR 25394.

Dave


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25125