[Bug fortran/48692] [4.7 Regression] ICE with gfortran.dg/module_write_1.f90

[burnus at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48692

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.0

--- Comment #2 from Tobias Burnus <burnus at gcc dot gnu.org> 2011-04-19 21:54:03 UTC ---
The symbol which is still referenced in "changed_syms" is the attr.generic
FL_PROCEDURE "len", which is the only symbol (->tlink == NULL).


In principle, the symbols should be committed via gfc_parse_file's:

    case ST_MODULE:
      push_state (&s, COMP_MODULE, gfc_new_block);
      accept_statement (st); /* This commits the pending symbols.  */
      [...]
      parse_module ();  /* Ditto, calls accept_statement.  */

That also nicely works - the symbol only pops up suddenly when calling
gfc_dump_module, which seems to be a bug in gfc_dump_module.

The reason that the issue was not seen before is that
"gfc_generate_module_code" calls calling gfc_generate_module_vars, which calls
gfc_trans_common which calls gfc_commit_symbols ().

Thus, a patch which papers the issue would add a call to "gfc_commit_symbols
();". However, the proper solutions to find out why gfc_dump_module leaves a
dangling symbol.