Thanks, I added that test case in r303642. I was struggling to identify the
behavior change.
On Mon, May 22, 2017 at 4:37 PM, Richard Smith <rich...@metafoo.co.uk>
wrote:
> On 22 May 2017 at 15:22, Reid Kleckner via cfe-commits <
> cfe-commits@lists.llvm.org> wrote:
>
>> On Mon, May 22, 2017 at 3:17 PM, Richard Smith <rich...@metafoo.co.uk>
>> wrote:
>>
>>> // If the StrTok is "eod", then it wasn't present. Otherwise, it
>>>> must be a
>>>> // string followed by eod.
>>>> - if (StrTok.is(tok::eod))
>>>> - ; // ok
>>>> - else if (StrTok.isNot(tok::string_literal)) {
>>>> + if (StrTok.is(tok::eod)) {
>>>> + // Treat this like "#line NN", which doesn't change file
>>>> characteristics.
>>>> + FileKind = SourceMgr.getFileCharacteristi
>>>> c(DigitTok.getLocation());
>>>>
>>>
>>> This change for "# <number>" handling makes sense (and I've checked and
>>> it matches GCC), but it looks like we don't have test coverage for either
>>> the old or new behavior. Can I interest you in adding some? :)
>>>
>>
>> We do have coverage for it, see test/Preprocessor/line-directive.c:
>> """
>> # 192 "glomp.h" 3 // System header.
>> typedef int y; // ok
>> typedef int y; // ok
>>
>> typedef int q; // q is in system header.
>>
>> #line 42 "blonk.h" // doesn't change system headerness.
>>
>> typedef int z; // ok
>> typedef int z; // ok
>>
>> # 97 // doesn't change system headerness.
>>
>> typedef int z1; // ok
>> typedef int z1; // ok
>> """
>>
>> We were just getting similar behavior because of the code I removed that
>> sends us to the #line directive handling when the filename isn't present.
>>
>
> Ah, I see. It looks like, prior to your change, we got self-inconsistent
> results. "# <number>" worked properly when compiling (and we do have
> coverage for that, per the above test), but not when preprocessing (for
> which we don't have test coverage). So for instance, this input:
>
> # 1 "foo" 3
> # 2
> x
>
> preprocesses as
>
> # 1 "foo" 3
> # 2 "foo"
> x
>
> before your change and as
>
> # 1 "foo" 3
> # 2 "foo" 3
> x
>
> after your change.
>
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
