[Bug fortran/53478] New: gfortran segfaults when module name clashes with C binding name of procedure

[Bug fortran/53478] New: gfortran segfaults when module name clashes with C binding name of procedure

[solomon.gibbs.lists at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53478

             Bug #: 53478
           Summary: gfortran segfaults when module name clashes with C
                    binding name of procedure
    Classification: Unclassified
           Product: gcc
           Version: 4.6.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: solomon.gibbs.lists@gmail.com


gfortran segfaults rather than producing a compiler error when a module
procedure is bound with the same name as the module.

e.g.

module exports
    implicit none

    contains

    subroutine f_exports() bind(C, name='exports')
    end subroutine

end module exports


gfortran -funderscoring -O0 -g -Wall -c -fmessage-length=0 -o "exports.o"
"../exports.f08"
f951.exe: internal compiler error: Segmentation fault

Possibly related to Bug 48858 ?

[Bug fortran/53478] gfortran segfaults when module name clashes with C binding name of procedure

[kargl at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53478

kargl at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2012-05-24
                 CC|                            |kargl at gcc dot gnu.org
     Ever Confirmed|0                           |1

--- Comment #1 from kargl at gcc dot gnu.org 2012-05-24 17:23:50 UTC ---
Confirmed.
(gdb) bt
#0  0x0000000202c0103c in strcmp () from /lib/libc.so.7
#1  0x000000000050aa91 in gfc_verify_binding_labels (sym=0x203cd1280)
    at ../../gcc4x/gcc/fortran/resolve.c:9847
#2  0x000000000052cf40 in do_traverse_symtree (st=0x203c183a0, st_func=0, 
    sym_func=0x50a840 <gfc_verify_binding_labels>)
    at ../../gcc4x/gcc/fortran/symbol.c:3386
#3  0x000000000051c030 in resolve_types (ns=0x203c63600)
    at ../../gcc4x/gcc/fortran/resolve.c:13984
#4  0x000000000051bef8 in resolve_types (ns=0x203c62c00)
    at ../../gcc4x/gcc/fortran/resolve.c:13965
#5  0x000000000050feec in gfc_resolve (ns=0xfffffffffffffff8)
    at ../../gcc4x/gcc/fortran/resolve.c:14054
#6  0x0000000000502a48 in gfc_parse_file ()
    at ../../gcc4x/gcc/fortran/parse.c:4594
#7  0x000000000053c90e in gfc_be_parse_file ()
    at ../../gcc4x/gcc/fortran/f95-lang.c:191
#8  0x00000000008f57f0 in compile_file () at ../../gcc4x/gcc/toplev.c:552
#9  do_compile () at ../../gcc4x/gcc/toplev.c:1874
#10 0x00000000008f5e4c in toplev_main (argc=2, argv=0x7fffffffd4e0)
    at ../../gcc4x/gcc/toplev.c:1950
#11 0x0000000000499a6d in _start ()
(gdb) up 1
#1  0x000000000050aa91 in gfc_verify_binding_labels (sym=0x203cd1280)
    at ../../gcc4x/gcc/fortran/resolve.c:9847
9847              if (sym->attr.if_source == IFSRC_DECL 
(gdb) list
9842
9843          bind_c_sym = gfc_find_gsymbol (gfc_gsym_root,
sym->binding_label);
9844          if (bind_c_sym != NULL 
9845              && strcmp (bind_c_sym->name, sym->binding_label) == 0)
9846            {
9847              if (sym->attr.if_source == IFSRC_DECL 
9848                  && (bind_c_sym->type != GSYM_SUBROUTINE 
9849                      && bind_c_sym->type != GSYM_FUNCTION) 
9850                  && ((sym->attr.contained == 1 
9851                       && strcmp (bind_c_sym->sym_name, sym->name) != 0) 
(gdb) print *bind_c_sym
$1 = {priority = 41101, left = 0x0, right = 0x0, name = 0x20400bf80 "exports", 
  sym_name = 0x0, mod_name = 0x0, binding_label = 0x0, type = GSYM_MODULE, 
  defined = 1, used = 0, where = {nextc = 0x203ca61f8, lb = 0x203ca61a0}, 
  ns = 0x203c62c00}

Note, sym->binding_label is 0x0

[Bug fortran/53478] gfortran segfaults when module name clashes with C binding name of procedure

[sgk at troutmask dot apl.washington.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53478

--- Comment #2 from Steve Kargl <sgk at troutmask dot apl.washington.edu> 2012-05-24 17:57:26 UTC ---
On Thu, May 24, 2012 at 05:23:50PM +0000, kargl at gcc dot gnu.org wrote:
> #1  0x000000000050aa91 in gfc_verify_binding_labels (sym=0x203cd1280)
>     at ../../gcc4x/gcc/fortran/resolve.c:9847
> 9847              if (sym->attr.if_source == IFSRC_DECL 
> (gdb) list
> 9842
> 9843          bind_c_sym = gfc_find_gsymbol (gfc_gsym_root,
> sym->binding_label);
> 9844          if (bind_c_sym != NULL 
> 9845              && strcmp (bind_c_sym->name, sym->binding_label) == 0)
> 9846            {
> 9847              if (sym->attr.if_source == IFSRC_DECL 
> 9848                  && (bind_c_sym->type != GSYM_SUBROUTINE 
> 9849                      && bind_c_sym->type != GSYM_FUNCTION) 
> 9850                  && ((sym->attr.contained == 1 
> 9851                       && strcmp (bind_c_sym->sym_name, sym->name) != 0) 
> (gdb) print *bind_c_sym
> $1 = {priority = 41101, left = 0x0, right = 0x0, name = 0x20400bf80 "exports", 
>   sym_name = 0x0, mod_name = 0x0, binding_label = 0x0, type = GSYM_MODULE, 
>   defined = 1, used = 0, where = {nextc = 0x203ca61f8, lb = 0x203ca61a0}, 
>   ns = 0x203c62c00}
> 
> Note, sym->binding_label is 0x0


This patch allows the code to compile.  I haven't
tried to construct a longer example that may show
that the module name  may collide with the C binding
name.


Index: resolve.c
===================================================================
--- resolve.c   (revision 187464)
+++ resolve.c   (working copy)
@@ -9848,6 +9848,8 @@ gfc_verify_binding_labels (gfc_symbol *s
               && (bind_c_sym->type != GSYM_SUBROUTINE 
                   && bind_c_sym->type != GSYM_FUNCTION) 
               && ((sym->attr.contained == 1 
+                  && bind_c_sym->sym_name != NULL
+                  && sym->name != NULL
                    && strcmp (bind_c_sym->sym_name, sym->name) != 0) 
                   || (sym->attr.use_assoc == 1 
                       && (strcmp (bind_c_sym->mod_name, sym->module) != 0))))

[Bug fortran/53478] gfortran segfaults when module name clashes with C binding name of procedure

[sgk at troutmask dot apl.washington.edu]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53478

--- Comment #3 from Steve Kargl <sgk at troutmask dot apl.washington.edu> 2012-05-24 18:44:54 UTC ---
On Thu, May 24, 2012 at 05:57:26PM +0000, sgk at troutmask dot
apl.washington.edu wrote:
> 
> This patch allows the code to compile.  I haven't
> tried to construct a longer example that may show
> that the module name  may collide with the C binding
> name.
> 

I've now constructed a simple test and needed to
revise the patch.

troutmask:sgk[263] cat a.f90
module exports

   implicit none

   contains

      subroutine f_exports() bind(C, name='exports')
         print *, 'Exported'
      end subroutine

end module exports
troutmask:sgk[264]  cat e.c
#include <stdio.h>

void exports(void);

int
main(void)
{
  exports();
  return 0;
}
troutmask:sgk[265] gfc4x -c a.f90
troutmask:sgk[266] cc -o z e.c a.o -L$HOME/work/4x/lib -lgfortran
troutmask:sgk[267] ./z
 Exported

The patch is attached.

[Bug fortran/53478] gfortran segfaults when module name clashes with C binding name of procedure

[dominiq at lps dot ens.fr]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53478

--- Comment #4 from Dominique d'Humieres <dominiq at lps dot ens.fr> 2012-06-25 12:05:24 UTC ---
I have the patch in comment #2 in my tree since some time now and it works as
advertised without regression. However, If I have understood Ian Harvey's
answer in comp.lang.fortran

https://groups.google.com/forum/?fromgroups#!topic/comp.lang.fortran/o_hapvMHxIE

the test in comment #3 is invalid. Is this correct?

[Bug fortran/53478] gfortran segfaults when module name clashes with C binding name of procedure

[aron at ahmadia dot net]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53478

Aron Ahmadia <aron at ahmadia dot net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |aron at ahmadia dot net

--- Comment #5 from Aron Ahmadia <aron at ahmadia dot net> 2013-04-15 10:06:20 UTC ---
I was able to reproduce this on 4.7.2 using Solomon's code snippet.  Is there
anything else I do to help out here?

[Bug fortran/53478] gfortran segfaults when module name clashes with C binding name of procedure

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53478

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu.org

--- Comment #6 from Tobias Burnus <burnus at gcc dot gnu.org> 2013-04-15 11:01:10 UTC ---
(In reply to comment #4)
> I have the patch in comment #2 in my tree since some time now and it works as
> advertised without regression. However, If I have understood Ian Harvey's
> answer in comp.lang.fortran
> https://groups.google.com/forum/?fromgroups#!topic/comp.lang.fortran/o_hapvMHxIE
> the test in comment #3 is invalid. Is this correct?

The test case in comment 0 and comment 3 are invalid in Fortran 2003; I think
they are valid in Fortran 2008 (cf. PR 48858 comment 9). However, I need a
quiet moment to disentangle the standard speak and to create a patch.

[Bug fortran/53478] gfortran segfaults when module name clashes with C binding name of procedure

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53478

Bug 53478 depends on bug 48858, which changed state.

Bug 48858 Summary: Incorrect error for same binding label on two generic interface specifics
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48858

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |FIXED