On 12 May 2014 22:24, Joseph S. Myers <[email protected]> wrote:
> On Mon, 12 May 2014, Manuel López-Ibáñez wrote:
>
>> I will be very surprised if the common defaults are overriding a FE
>> default and it is not a bug in the FE.
>
> Well, I think that needs justification, not just "very surprised".
> Identify (presumably in an automated way) all the cases where an option is
> handled by more than one of the handlers and, for each one, describe what
> cases would be affected by a change of order and why the proposed new
> order gives the desired effects.
common_handle_option_auto handles just OPT_Wextra,
OPT_Wuninitialized: and OPT_Wunused.
Fortran is the only one that handles OPT_Wextra in the FE, and the
current order prevents Fortran to override the current default, while
the new order will allow it.
These are the options handled in gcc/opts.c
OPT_auxbase_strip
OPT_aux_info
OPT_d
OPT_fcall_saved_
OPT_fcall_used_
OPT_fdbg_cnt_
OPT_fdbg_cnt_list
OPT_fdebug_prefix_map_
OPT_fdiagnostics_color_
OPT_fdiagnostics_show_caret
OPT_fdiagnostics_show_location_
OPT_fdiagnostics_show_option
OPT_fdump_
OPT_ffast_math
OPT_ffixed_
OPT_finline_limit_
OPT_finstrument_functions_exclude_file_list_
OPT_finstrument_functions_exclude_function_list_
OPT_flto
OPT_fmax_errors_
OPT_fmessage_length_
OPT_fopt_info
OPT_fopt_info_
OPT_fpack_struct_
OPT_fplugin_
OPT_fplugin_arg_
OPT_fprofile_generate
OPT_fprofile_generate_
OPT_fprofile_use
OPT_fprofile_use_
OPT_frandom_seed
OPT_frandom_seed_
OPT_fsanitize_
OPT_fsched_stalled_insns_
OPT_fsched_stalled_insns_dep_
OPT_fsched_verbose_
OPT_fshow_column
OPT_fstack_check_
OPT_fstack_limit
OPT_fstack_limit_register_
OPT_fstack_limit_symbol_
OPT_fstack_usage
OPT_ftrapv
OPT_ftree_vectorize
OPT_funsafe_math_optimizations
OPT_fuse_ld_bfd
OPT_fuse_ld_gold
OPT_fuse_linker_plugin
OPT_fwrapv
OPT_g
OPT_gcoff
OPT_gdwarf
OPT_gdwarf_
OPT_ggdb
OPT_gsplit_dwarf
OPT_gstabs
OPT_gstabs_
OPT_gvms
OPT_gxcoff
OPT_gxcoff_
OPT__help
OPT__help_
OPT_O
OPT_Ofast
OPT_Og
OPT_Os
OPT__param
OPT_pedantic_errors
OPT__target_help
OPT__version
OPT_w
OPT_Werror
OPT_Werror_
OPT_Wfatal_errors
OPT_Wframe_larger_than_
OPT_Wlarger_than_
OPT_Wstack_usage_
OPT_Wstrict_aliasing
OPT_Wstrict_overflow
OPT_Wsystem_headers
I found the ones handled in the FEs by using:
$ for o in $(grep 'case OPT_' gcc/opts.c | sed 's/^\s\+case \+//g' |
sort -u); do grep $o gcc/ada/gcc-interface/misc.c
gcc/fortran/options.c gcc/fortran/cpp.c gcc/java/lang.c
gcc/lto/lto-lang.c gcc/c-family/c-opts.c; done
gcc/fortran/options.c: case OPT_d:
gcc/fortran/options.c: case OPT_d:
gcc/fortran/cpp.c: case OPT_d:
gcc/c-family/c-opts.c: case OPT_d:
gcc/java/lang.c: case OPT_fdump_:
Of these, OPT_fdump_ is deferred in gcc/opts.c, so the order does not matter.
For OPT_d, the common handler does:
case 'D': /* These are handled by the preprocessor. */
case 'I':
case 'M':
case 'N':
case 'U':
break;
While Fortran does:
case OPT_d:
for ( ; *arg; ++arg)
switch (*arg)
{
case 'D':
case 'M':
case 'N':
case 'U':
gfc_cpp_option.dump_macros = *arg;
break;
case 'I':
gfc_cpp_option.dump_includes = 1;
break;
}
break;
and C/C++ does:
while ((c = *arg++) != '\0')
switch (c)
{
case 'M': /* Dump macros only. */
case 'N': /* Dump names. */
case 'D': /* Dump definitions. */
case 'U': /* Dump used macros. */
flag_dump_macros = c;
break;
case 'I':
flag_dump_includes = 1;
break;
}
Am I missing something?
Cheers,
Manuel.
