This is the mail archive of the
mailing list for the GCC project.
Re: RFA: Remove TARGET_SWITCHES from the SH port
Joern RENNECKE <firstname.lastname@example.org> writes:
> Richard Sandiford wrote:
>>OK, thanks for the explanation. In that case, would it be OK to commit
>>the patch without the removal of superh.h and superh64.h? The patch
>>won't really break anything related to them, in the sense that defining
>>TARGET_PROCESSOR_SWITCHES wouldn't have any effect in the current
>>sources anyway, even if the header files _were_ used. The macros
>>will then just document for what a future .opt file would need to contain.
> Yes, except that one of the messages in your patch is not quite right:
>> + if ((old_flags ^ target_flags) & (MASK_SH1 | MASK_SH2 | MASK_SH3
>> + | MASK_SH_E | MASK_HARD_SH4
>> + | MASK_FPU_SINGLE | MASK_SH4))
>> + return _("created and used with different architectures");
> I think this should rather say:
> "created and used with different architectures / ABIs" . -m4, -m4-single
> and -m4-single-only are for the same architecture, but different ABIs.
OK, thanks, installed with that change.
> FWIW the flag_mask mechanism was written with the intention that it'd fit
> into toplev.c (given a suitable interface to supply the mask). But as the
> SH port is still the only one to redefine TARGET_PCH_VALID_P, there is
> probably not much demand for this.
FWIW, I'm thinking about ways in which the options scripts could
help here. No promises though.