[Patch] Fix for bug 15246

[Eli Bendersky]

On Wed, Feb 13, 2013 at 9:56 AM, Jyotsna Verma <jverma at codeaurora.org> wrote:
>>So if it produces an incorrect enum declaration, can't this be detected in
> a
>>test? Or are you trying to say that the bug is in a specific TG backend
> that's not
>>currently being tested? Would this be difficult to implement?
>
> Yes, the bug is specific to a TG backend, CodeGenMapTable.cpp. All the
> current tests for TableGen appear to be for the frontend/parser. I don't
> think we have any backend specific test at this point. In order to have such
> a test, it would require some basic infrastructure in a test .td file,
> mostly the classes defined in include/llvm/Target/Target.td. This may result
> into a test with a couple of hundreds of lines.

Or just include the needed .td file(s)?

> I'm not sure if it's
> desirable just to test a trivial functionality of the backend. It will be
> more appropriate when used for thorough testing.
>
> If it's oaky, I would first like to push the proposed fix for the TG backend
> (without the Hexagon change). In the meantime, I will see if I can come up
> with an independent test to validate the fix.

Splitting the patches is a good idea in general. You should get a LGTM
from someone else, as I'm not the code owner in these parts.

Eli