Hi Renato,

On Wed, Dec 11, 2013 at 4:34 PM, Renato Golin <[email protected]> wrote:
>
> I think the problem here is that there isn't a single line in Clang (I
> could find) that uses that information, so it's impossible to create a
> test that doesn't involve external code.
>
> If we had a dumper that used this code in some standard way, we could
> use that as both to test and as a documentation on how to use it (for
> the external users).

John Thompson built a tool like that a few weeks back:
https://github.com/llvm-mirror/clang-tools-extra/tree/master/pp-trace

for the specific purpose of testing PPCallbacks. He's been trying to
call attention to getting it into Clang core.

I'm still on the fence whether it's a good idea to build a separate
tool to test PPCallbacks, because failures would signal problems with
the tool, not with PPCallbacks. Intuitively, I think unit tests would
be better, but that was discouraged by someone else in the original
review.

FWIW,
- Kim
_______________________________________________
cfe-commits mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits

Reply via email to