Re: [Patch, fortran] PR30476 - [Regression 4.2, 4.3] Via other module imported generic interface rejected

[Paul Thomas <paulthomas2 at wanadoo dot fr>]

Hi Tobias,
> /home/tob/projects/gcc/gcc/fortran/module.c:3100: Warning: the address
> of "module" will always evaluate as "true"
>
> module is defined as:
> char module[GFC_MAX_SYMBOL_LEN + 1];
>
> Maybe it is better to use   if (module[0] && ...  ?
No, it's better to eliminate that part of the condition completely; it's 
pure reflex to check that a pointer is there.  Thanks

This part of the code now reads:

       {
         /* Unless sym is a generic interface, this reference
        is ambiguous.  */
         gfc_symtree *st;
         p = p ? p : name;
         st = gfc_find_symtree (gfc_current_ns->sym_root, p);
         if (!sym->attr.generic
           && sym->module != NULL
           && strcmp(module, sym->module) != 0)
       st->ambiguous = 1;
       }

> Is my assumption correct that  st->ambiguous gets already set correctly
> via gfc_find_symtree?
No, st->ambiguous only gets set in two places; both in module.c.
> I build & regtested it on openSUSE/x86_64 with no new regressions.
Is it OK with the above change?

Paul

PS reply:
> Post scriptum and unrelated to this patch:
>
> For some reasons I get a core-dump when running interface_10.f90 and
> somehow two ambiguous errors are not detected when compiling 
> interface_11.f90 (dg-error in lines 72 and 88). However, I also have
> these failures without the patch applied.
You answered your own question on this one :-)
> * c_by_val_1 and vect/vect-4.f90 fail on i64:
>   http://gcc.gnu.org/ml/gcc-testresults/2007-01/msg00633.html
>   http://gcc.gnu.org/ml/gcc-testresults/2007-01/msg00612.html
The c_by_val bit is PR30432 which I am in no position to do anything 
with because I do not see it. Andrew Pinski has raised the concern, in 
the PR thread, that this not a gfortran issue.
> * open_errors.f90, string_0xfe_0xff_1.f90 and vect/vect-5.f90 on
> sparc-sun-solaris
>   http://gcc.gnu.org/ml/gcc-testresults/2007-01/msg00628.html
>
> * on aix5 c_by_val, static_linking_1, string_0xfe_0xff_1.f90, value_4 and
>   gfortran.fortran-torture/execute/intrinsic_nearest.f90 fail
>   http://gcc.gnu.org/ml/gcc-testresults/2007-01/msg00624.html
ditto, I suspect.
> * on hppa-linux fail bound_2, cray_pointers_2, and string_0xfe_0xff_1.f90
>   http://gcc.gnu.org/ml/gcc-testresults/2007-01/msg00622.html
>
> * on another x86_64 (-m32 and -m64) fails string_0xfe_0xff_1.f90
>   http://gcc.gnu.org/ml/gcc-testresults/2007-01/msg00613.html
I think that Thomas has dealt with the last?