[Bug fortran/53552] Add support for !GCC$ warning directive (cf. #pragma warning)

["burnus at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=53552

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

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

--- Comment #2 from Tobias Burnus <burnus at gcc dot gnu.org> 2012-06-04 13:38:51 UTC ---
(In reply to comment #1)
> Wouldn't it be better to make Fortran use the common GCC diagnostics machinery?

In some off-line discussion, Manuel wrote:
| diagnostics-core.h implements that core
| entry points. A FE needs to implement their own printf-like codes (see
| tree-diagnostics.c) and pretty-printer of expressions (similar to
| c-pretty-printer.c but for its own AST).

Regarding the multi-location support of gfortran:

| this is not implemented. But since the FEs need to handle
| their own format specifiers, it is a matter of interfacing your
| current implementation with the common diagnostics machinery.

| you could start by simply adding an interface to the current
| Fortran diagnostics machinery to access it from diagnostics.c. That
| is, implement enough glue code so you can call the current Fortran
| code passing through warning_at/error_at (or just re-wire gfc_warning
| to pass through diagnostic_report_diagnostic). This will give you
| #pragma GCC diagnostics, Werror=, and fdiagnostics-show-option almost
| for free.


I think as part of the implementation, one should pass to gfc_warning the flag
which triggers it (or 0 for default warning), similar to "warning_at"; example:
  warning_at (loc, OPT_Wcast_qual, ...

For hooking up front-end warnings, maybe java/lang.c can be of some help. Or
cp/error.c (esp. cp_print_error_function and cp_printer).