> (a) getSubArch would decouple it from -march, which IMO makes sense to > also recognize. If you request ARMv7 or ARMv8 as codegen target, it > doesn't make sense to produce a binary that can't run on it.
Fair point. Please add a test for -march too then, so no overzealous refactorer like me can come along in future and break it. > As I said, I don't know what Linux wants, so hooking it up has to be > done by someone who at least can test the correctness. And as I've said, I've investigated Linux behaviour and think it's what's been implemented here. I'm confident we want this behaviour for the Linux operating system and pure -eabi triples. I think both of those will be accomplished by adding it to gnutools::Link. Tim. _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
