[cfe-dev] ##

[Eli Friedman]

On Sun, May 24, 2009 at 4:34 AM, Sascha Wildner <swildner at erpicon.de> wrote:
> Eli Friedman schrieb:
>>
>> On Sun, May 24, 2009 at 2:55 AM, Sascha Wildner <swildner at erpicon.de>
>> wrote:
>>>
>>> Hi all,
>>>
>>> I've noted a difference between gcc and clang using this test:
>>
>> Is the difference significant to some program using clang's
>> preprocessor output?  If so, please file a bug; otherwise, it doesn't
>> really matter.  Making our -E output be byte-for-byte identical with
>> gcc's output isn't very important.
>
> Yes, it affects some assembler code which is generated by macros when
> building DragonFly BSD which uses __CONCAT(.,blah). I'll file a report for
> that.

Okay, I just fixed it in r72370.

> Note that it does work, when "a##b" is used instead of "a ## b". And it does
> not generate an extra space in clang if 'a' isn't '.'.

Concatenating two identifiers discards the RHS token; concatenating a
period with an identifier gets handled significantly differently.

-Eli