[PATCH] D14568: Handle ARMv6KZ naming

[Renato Golin via llvm-commits]

rengolin accepted this revision.
rengolin added a comment.
This revision is now accepted and ready to land.

LGTM, thanks!


================
Comment at: unittests/ADT/TripleTest.cpp:859
@@ -858,3 +858,3 @@
     llvm::Triple Triple("armv6k-unknown-eabi");
-    EXPECT_EQ("arm1176jzf-s", Triple.getARMCPUForArch());
+    EXPECT_EQ("arm1176j-s", Triple.getARMCPUForArch());
   }
----------------
tyomitch wrote:
> rengolin wrote:
> > This example came from somewhere I can't remember... Maybe FreeBSD? Or RPi?
> This comes from your patch from half a year ago (r237797) which changed the triple being tested from "armv6-unknown-freebsd" to "armv6k-unknown-eabi".
> 
> I suspect you were thus recording the existing state of the target parser, rather than an intention behind it?
> 
> I'm not sure it makes a lot of sense to have an ARMv6KZ core the default CPU for both ARMv6K and ARMv6KZ.
> If one wants to target ARMv6KZ, they would specify ARMv6KZ as the target, wouldn't they?
> If one wants to target ARMv6KZ, they would specify ARMv6KZ as the target, wouldn't they?

The problem here is how systems *do* reference those architectures and not how they *should*.

However, I'm willing to sacrifice legacy in name of clarity. In the long run, it's probably cheaper to deal with a few backslashes than carry legacy forever for no apparent reason.


http://reviews.llvm.org/D14568