This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: pr29759


On Fri, Nov 10, 2006 at 10:44:57AM +0100, Daniel Franke wrote:
> Great, gmail removed the attachment ...
> Here it is.
> 
> 2006/11/10, Daniel Franke <franke.daniel@gmail.com>:
> >2006/11/10, Steve Kargl <sgk@troutmask.apl.washington.edu>:
> >> On Fri, Nov 10, 2006 at 01:17:49AM +0100, Daniel Franke wrote:
> >> >
> >> > please find attached a proposed fix for PR29759 (ICE on OMP line
> >> > continuation).
> >> >
> >> > The ICE occured because all lines starting with "!$" where checked
> >> > against "!$OMP", even if they were known OMP conditional lines. The
> >> > fix for this should qualify as obvious. Is a testcase needed?

Actually, it is not obvious.  IMHO the testcase is invalid, OpenMP 2.5
in 2.1 says that:
"Directives are case-insensitive. Directives cannot be embedded within continued
statements, and statements cannot be embedded within directives."
In
!$OMP PARALLEL &
!$ NUM_THREADS(2)
!$OMP END PARALLEL
NUM_THREADS(2) is a statement, not directive.
So we need to issue diagnostics, not silently ignore it.
I'm not sure if
!$OMP PARALLEL &
! This is a comment
!$OMP NUM_THREAD(2)
!$OMP END PARALLEL
is valid or not though.

> >Regtested on i686-pc-linux-gnu as
> >$> make check-gfortran RUNTESTFLAGS="gomp.exp"

All OpenMP Fortran only changes need to be tested by both
make check-gfortran RUNTESTFLAGS="gomp.exp"
and
make -C ../*/libgomp check

	Jakub


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]