This is the mail archive of the 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: Revision 162491 -- fortran -fwhole-file regressions

Dominique Dhumieres wrote:
> However if it was done to expose the feature to a broader audience and
> you are planning to iron out the last wrinkles, I think the commit
> can be kept in place. The failing tests on i386-*-freebsd can use
> a dg-option -fno-whole-file to temporarily silence the failures.

There are two reasons: I think the patch is ready and I want to expose
it to a wider audience.

Steve Kargl wrote:
> There are 3 choices.  1) Fix the regressions.  2) Back out
> the commit. 3) Wait 48 hours, and I'll back out the commit
> myself.

How about letting me wake up first before threatening to back out
commits. In general I also need help if there are failures on platforms,
which I cannot access. But in this case, doing -ffloat-store seems to be
enough ;-) By the way, the inlining shows that gfortran had a missed
optimization :-)

The next thing which has to be fixed is the single-file CP2K - the
normal multi-file CP2K seems to work, though. But reducing a 304361 LOC
program takes a while - OK, Joost reduced it to 60498 LOC, but that's
still too much to look at the dumps. (Cf. PR 40011.)


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