[Bug fortran/84394] compiler error when using modules with derived types in block data subprograms

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84394

--- Comment #2 from kargl at gcc dot gnu.org ---
(In reply to kargl from comment #1)

> Breakpoint 1, check_conflict (attr=attr@entry=0x203caec38, 
> name=0x203c22430 "_deallocate", where=where@entry=0x203caec78)
> at ../../gcc/gcc/fortran/symbol.c:485
> 485   gfc_error("%s attribute not allowed in BLOCK DATA program "
> (gdb) p *where
> $1 = {nextc = 0x0, lb = 0x0}
> 
> So, gfortran is inserting a _deallocate (perhaps for a temporary array).
> I suspect were can the conflict check for a name starting with an
> underscore.

I forgot to mention that attr->access = ACCESS_PRIVATE, which
the actual root cause.  I also suspect that the attr is re-using
a previous attr struct without clear it first.   In any case, the
following works around the problem.


Index: symbol.c
===
--- symbol.c(revision 257665)
+++ symbol.c(working copy)
@@ -479,11 +479,10 @@ check_conflict (symbol_attribute *attr, const char *na
   if (attr->intent != INTENT_UNKNOWN)
a1 = intent;

-  if (a1 != NULL)
+  if (a1 != NULL && *name != '_')
{
- gfc_error
-   ("%s attribute not allowed in BLOCK DATA program unit at %L",
-a1, where);
+ gfc_error("%s attribute not allowed in BLOCK DATA program "
+   "unit at %L", a1, where);
  return false;
}
 }

[Bug fortran/84394] compiler error when using modules with derived types in block data subprograms

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84394

kargl at gcc dot gnu.org changed:

   What|Removed |Added

 Status|UNCONFIRMED |NEW
   Last reconfirmed||2018-02-15
 CC||kargl at gcc dot gnu.org
 Ever confirmed|0   |1

--- Comment #1 from kargl at gcc dot gnu.org ---
Further reduced testcase 

module mod1
   implicit none
   type :: type1
  integer :: i1
   end type type1
end module

module mod2
   implicit none
   contains
  subroutine sub1
 integer vals
 common /block1/ vals(5)
 if (any(vals /= [1, 2, 3, 4, 5])) stop 1
  end subroutine
end module

block data blkdat
  use mod1
  integer vals
  common /block1/ vals(5)
  data vals/1, 2, 3, 4, 5/
end block data blkdat

program main
  use mod2, only: sub1
  implicit none
  call sub1
end program

Looking at
0x7f911f check_conflict
../../gcc/fortran/symbol.c:485

This is really odd.  gdb shows 

Breakpoint 1, check_conflict (attr=attr@entry=0x203caec38, 
name=0x203c22430 "_deallocate", where=where@entry=0x203caec78)
at ../../gcc/gcc/fortran/symbol.c:485
485   gfc_error("%s attribute not allowed in BLOCK DATA program "
(gdb) p *where
$1 = {nextc = 0x0, lb = 0x0}

So, gfortran is inserting a _deallocate (perhaps for a temporary array).
I suspect were can the conflict check for a name starting with an
underscore.

[Bug fortran/84394] compiler error when using modules with derived types in block data subprograms

[kargl at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84394

kargl at gcc dot gnu.org changed:

   What|Removed |Added

   Priority|P3  |P4
  Known to fail||7.3.1, 8.0