This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Fortran regressions with revision 116758
- From: Andrew Pinski <pinskia at physics dot uc dot edu>
- To: hjl at lucon dot org (H. J. Lu)
- Cc: pinskia at physics dot uc dot edu (Andrew Pinski), gcc at gcc dot gnu dot org, kargls at comcast dot net
- Date: Thu, 7 Sep 2006 17:20:27 -0400 (EDT)
- Subject: Re: Fortran regressions with revision 116758
>
> On Thu, Sep 07, 2006 at 05:03:35PM -0400, Andrew Pinski wrote:
> > >
> > > I got failures like
> > >
> > > compiler exited with status 1
> > > output is:
> > > In file
> > > /net/gnu-13/export/gnu/src/gcc/gcc/libgomp/testsuite/libgomp.fortran/reduction3.f90:20^M
> > > ^M
> > > !$ if (i .ne. -2147483648 .or. any (ia .ne. -2147483648)) v = .true.^M
> > > 1^M
> > > Error: Integer too big for its kind at (1)^M
> > >
> > > FAIL: libgomp.fortran/reduction3.f90 -O0 (test for excess errors)
> >
> > This is a bug in the testcase and not in Gfrotran.
> > The reason is 2147483648 is parsed seperate from - just like in C.
>
> The C/C++ compilers in gcc 4.2 revision 116758 have no problems with
> -2147483648.
But you should get a warning with a 32bit target saying that it is unsigned.
-- Pinski