[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[dje at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #15 from David Edelsohn  ---
The AIX failures are fixed.

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #14 from ro at CeBiTec dot Uni-Bielefeld.DE  ---
> --- Comment #12 from Jerry DeLisle  ---
> Rainer, please confirm the fix commited on trunk when you have time.

Sure: I had it in my tree during this weekend's bootstraps and all
failures are gone, as expected.

Thanks.
Rainer

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

Jakub Jelinek  changed:

   What|Removed |Added

 Status|ASSIGNED|RESOLVED
 CC||jakub at gcc dot gnu.org
 Resolution|--- |FIXED

--- Comment #13 from Jakub Jelinek  ---
Assuming it is fixed, please reopen if not.

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[jvdelisle at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #12 from Jerry DeLisle  ---
Rainer, please confirm the fix commited on trunk when you have time.

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[jvdelisle at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #11 from Jerry DeLisle  ---
Author: jvdelisle
Date: Fri Oct 21 18:02:32 2016
New Revision: 241422

URL: https://gcc.gnu.org/viewcvs?rev=241422=gcc=rev
Log:
2016-10-21  Jerry DeLisle  

PR libfortran/78055
* io/io.h (st_parameter_dt): Restore GFC_IO_INT to maintain
alignment.

Modified:
trunk/libgfortran/ChangeLog
trunk/libgfortran/io/io.h

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[jvdelisle at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

Jerry DeLisle  changed:

   What|Removed |Added

 Status|WAITING |ASSIGNED
   Assignee|unassigned at gcc dot gnu.org  |jvdelisle at gcc dot 
gnu.org

--- Comment #10 from Jerry DeLisle  ---
Working on it.

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[dje at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

David Edelsohn  changed:

   What|Removed |Added

 Target|sparc-sun-solaris2.12   |sparc-sun-solaris2.12,
   ||powerpc-ibm-aix*
 CC||dje at gcc dot gnu.org
   Host|sparc-sun-solaris2.12   |sparc-sun-solaris2.12,
   ||powerpc-ibm-aix*
  Build|sparc-sun-solaris2.12   |sparc-sun-solaris2.12,
   ||powerpc-ibm-aix*

--- Comment #9 from David Edelsohn  ---
FYI, I see the same failures on AIX.

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #8 from ro at CeBiTec dot Uni-Bielefeld.DE  ---
> --- Comment #7 from Jerry DeLisle  ---
> Would you please apply the patch from here and see if it fixes also or breaks
> more.
>
> https://gcc.gnu.org/ml/fortran/2016-10/msg00133.html

I just did, but with and without the patch for PR78055, I now get the
same set of failures that before were present for 32-bit only also for 64-bit.

Rainer

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[jvdelisle at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #7 from Jerry DeLisle  ---
Rainer,

Would you please apply the patch from here and see if it fixes also or breaks
more.

https://gcc.gnu.org/ml/fortran/2016-10/msg00133.html

Thanks,

Jerry

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[jvdelisle at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #6 from Jerry DeLisle  ---
I think I understand. The Frontend needs to access the parameters before and
after the dtp->u.p section. When deleting that integer, that section was
shortened, so the placement of the parameters after are offset. I am not sure
why it does not break everywhere.  I am thinking the patch to cleanup the dtp
stucture will realign everything.

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[jvdelisle at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #5 from Jerry DeLisle  ---
Ok I will commit that little chink for now. My next patch really slams this
structure so I think I will send it to you first to see what breaks.  In fact I
just pinged the post for it.

In the meantime need to figure out why we need the not used GFC_IO_INT.

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #4 from ro at CeBiTec dot Uni-Bielefeld.DE  ---
> --- Comment #3 from ro at CeBiTec dot Uni-Bielefeld.DE  Uni-Bielefeld.DE> ---
[...]
> Seems so, especially since sparc is more sensitive here.  I've just
> rebuilt libgfortran with the patch applied and manually ran the first
> failing test: it passes now.  Running a full make check-gfortran for
> both 32 and 64-bit multilibs now.

Completed successfully now and testresults are as they were before.

Thanks.
Rainer

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #3 from ro at CeBiTec dot Uni-Bielefeld.DE  ---
> --- Comment #2 from Jerry DeLisle  ---
> As a test, try this patch.
>
> diff --git a/libgfortran/io/io.h b/libgfortran/io/io.h
> index edc520a9..00ced533 100644
> --- a/libgfortran/io/io.h
> +++ b/libgfortran/io/io.h
> @@ -514,6 +514,7 @@ typedef struct st_parameter_dt
>  large enough to hold a complex value (two reals) of the
>  largest kind.  */
>   char value[32];
> + GFC_IO_INT not_used;
>   formatted_dtio fdtio_ptr;
>   unformatted_dtio ufdtio_ptr;
> } p;
>
> Maybe it is an alignment issue???

Seems so, especially since sparc is more sensitive here.  I've just
rebuilt libgfortran with the patch applied and manually ran the first
failing test: it passes now.  Running a full make check-gfortran for
both 32 and 64-bit multilibs now.

Thanks.
Rainer

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[jvdelisle at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #2 from Jerry DeLisle  ---
As a test, try this patch.

diff --git a/libgfortran/io/io.h b/libgfortran/io/io.h
index edc520a9..00ced533 100644
--- a/libgfortran/io/io.h
+++ b/libgfortran/io/io.h
@@ -514,6 +514,7 @@ typedef struct st_parameter_dt
 large enough to hold a complex value (two reals) of the
 largest kind.  */
  char value[32];
+ GFC_IO_INT not_used;
  formatted_dtio fdtio_ptr;
  unformatted_dtio ufdtio_ptr;
} p;

Maybe it is an alignment issue???

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[dominiq at lps dot ens.fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

Dominique d'Humieres  changed:

   What|Removed |Added

 Status|UNCONFIRMED |WAITING
   Last reconfirmed||2016-10-20
 Ever confirmed|0   |1

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[jvdelisle at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

--- Comment #1 from Jerry DeLisle  ---
Looking, but need more info.

[Bug libfortran/78055] [7 regression] Many new gfortran test failures

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=78055

Rainer Orth  changed:

   What|Removed |Added

   Target Milestone|--- |7.0