Depends on D48988.
Details
Diff Detail
- Repository
- rC Clang
Event Timeline
This seems reasonable if you need the support in the assembler. However, please add a test to ensure that the paths are mapped when invoking the assembler rather than the compiler.
I added the tests to check the mapping logic through llvm-mc in D48988. In this revision, I merely test if the driver is passing the flags to cc1as. The only thing that is untested is if cc1as_main is setting the options in MCContext correctly. If you want to see a test for that, please can you guide me a little on an appropriate way to test this within tools/clang/test.
That would be more of an integration test, which we don't generally have in the 'lit' tests. I think demonstrating correct mapping by llvm-mc is fine.
I have made a suggestion for the test in this patch, which helps show the option is being passed to cc1as as intended.
test/Driver/debug-prefix-map.S | ||
---|---|---|
4 | To show that the option is on the cc1as command line specifically: |
To show that the option is on the cc1as command line specifically:
CHECK: cc1as
CHECK-SAME: -fdebug-prefix-map=old=new