[Patch, fortran] PR20880 - Failure to detect procedures encompassing own interface

[Tobias Burnus]

Hi,

Paul Thomas wrote:
> Thanks to Erik and Tobias for picking up this fault in the patch and
> for helping me correct it.  Perhaps one of you can test it again, please?
Regression tested on x86_64-unknown-linux-gnu with glibc's
MALLOC_PERTURB_=B and MALLOC_CHECK_=2 set.
# of expected passes          15403
# of expected failures        7
# of unsupported tests        17

I also did a check using valgrind. The only, non-serious (and probably
not new) issue in gfortran is:
(checked with --leak-check=full using func_derived_3.f90 as test case):

==18838== 21,836 (3,176 direct, 18,660 indirect) bytes in 36 blocks are
definitely lost in loss record 5 of 9
==18838==    at 0x4C22889: malloc (in
/usr/lib64/valgrind/amd64-linux/vgpreload_memcheck.so)
==18838==    by 0x9536F7: xmalloc (xmalloc.c:147)
  PTR newmem;
  newmem = malloc (size); // <--- line 147
==18838==    by 0x439BD4: gfc_getmem (misc.c:40)
  void *p;
  p = xmalloc (n); // <--- line 40
==18838==    by 0x45A442: gfc_new_symbol (symbol.c:1987)
  gfc_symbol *p;
  p = gfc_getmem (sizeof (gfc_symbol)); // <--- line 1987
==18838==    by 0x45A57C: gfc_get_sym_tree (symbol.c:2122)
  gfc_symbol *p;
  p = gfc_new_symbol (name, ns); // <--- line 2122
==18838==    by 0x45A711: gfc_get_symbol (symbol.c:2173)
==18838==    by 0x411CF7: gfc_match_modproc (decl.c:4198)


> --- gcc/testsuite/gfortran.dg/interface_3.f90	(revision 0)
> + ! Contributed by
>   
By whom?

Tobias