> I really disagree with the direction of this patch. > > Currently we have succeeded at factoring all of the GCC '-f' flag-spelling > compatibility into the driver, and keeping it out of the frontend. I would > personally rather keep it that way, and have a much more constrained and > justified command line flag interface at the CC1 boundary. I find this > separation of concerns extremely useful.
If the -cc1 option handling can be made *really* simple, I would be ok with it. By really simple I mean something like "all boolean options default to false, the driver just passes a list of flags to enable". In any case, the main intention of this patch work is avoid the code duplication/bugs when handling a -f/-fno- pair, so it can be done in the driver too. The attached patch is the driver version of the previous one. The next steps are somewhat similar: Start a .td file with simple definitions for f flags. Use tablegen to produce an enum of all flags and matching code. The translation to -cc1 options and the logic in -cc1 itself would remain manual for now. Cheers, Rafael _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
