[Bug fortran/51073] _gfortran_caf_register incorrectly assumes malloc(0) returns non-NULL

["burnus at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=51073

--- Comment #1 from Tobias Burnus <burnus at gcc dot gnu.org> 2011-11-10 08:13:54 UTC ---
(In reply to comment #0)
> I am debugging why some of the gfortran tests are failing.  I have tracked NN
> failures down to this code in caf/mpi.c around line 155.

I think you are looking at the wrong file and the cause is due to
libgfortran/caf/single.c, which contains similar code.

With coarrays, gfortran can either use single.c ("libcaf_single.a") or mpi.c.
While "libcaf_single.a" is build on all systems and tested in
gcc/testsuite/gfortran.dg/coarrary/, mpi.c is not build at all. The reason is
that building mpi.c strongly depends on the used Message Passing Interface
implementation. Thus, it is currently left to the user to fetch mpi.c and
libcaf.h and run "mpicc -O2 -c mpi.c". Hence, it is also not automatically
tested.

I am considering to allow building mpi.c by passing some flags to configure. I
am also thinking of supporting it in the testsuite (with some environment
variables for the mpi-library path and the running command). However, I have
not yet done so.

>   /* Token contains only a list of pointers.  */
>   local = malloc (size);

> Upon successful completion with size not equal to 0, malloc() shall return a
> pointer to the allocated space. If size is 0, either a null pointer or a
> unique pointer that can be successfully passed to free() shall be returned.

Good point - I have think about how to fix that correctly.