[cfe-commits] r89100 - /cfe/trunk/lib/Frontend/FrontendOptions.cpp

[Nuno Lopes]

> On Sun, Nov 22, 2009 at 11:52 AM, Nuno Lopes <nunoplopes at sapo.pt> wrote:
>> Hi,
>>
>> Ok, so here is what I get:
>> vmware ~ # touch x.hpp
>> vmware ~ # gcc -fsyntax-only x.hpp
>> vmware ~ # g++ -fsyntax-only x.hpp
>> vmware ~ # gcc --version
>> gcc (Gentoo 4.3.4 p1.0, pie-10.1.5) 4.3.4
>> (...)
>>
>>
>> Gcc 4.4 behaves the same way in fedora 12..
>> You're right that clang doesn't still work, but at least this patch fixes
>> clang-cc (I just verified now both things with a clean tree). In fact, 
>> this
>> is one of the blockers of building inkscape.
>
> Ok, I fixed this in r89630. My main point was that clang-cc is
> irrelevant here, you didn't need to add it to clang-cc to fix the real
> problem.
>
> Unless you are one of those people trying to use clang-cc to actually
> do things, which I am trying to wean people off of. :)

Thanks for fixing it.
I'm actually trying stepping away from clang-cc these days, but sometimes 
the habit is stronger.. :)  Anyway, feel free to revert my previous patch, 
or let me know if you want me to do it, as apparently it is bogus. I simply 
don't know how I tested the patch and it worked before commiting it.. :P

Nuno


>> ----- Original Message -----
>>>
>>> Hi Nuno,
>>>
>>> On Tue, Nov 17, 2009 at 2:19 PM, Nuno Lopes <nunoplopes at sapo.pt> wrote:
>>>>
>>>> The problem was that if you run 'clang/clang-cc -fsyntax-only file.hpp'
>>>> the
>>>> file was not being interpreted as a C++ file. gcc, on the other hand,
>>>> does
>>>> interpret .hpp files correctly as c++.
>>>> I don't have the code here handy to be sure, but the problem I was 
>>>> facing
>>>> was something along these lines (which was breaking the detection of
>>>> boost
>>>> by inkscape's configure).
>>>
>>> Yes, I understood that, the problem is that (a) your change doesn't
>>> fix this, and (b) my gcc doesn't do this. Maybe this has changed in
>>> upstream gcc, but (a) remains.
>>> --
>>> ddunbar at giles:tmp$ touch x.hpp
>>> ddunbar at giles:tmp$ clang -fsyntax-only t.hpp
>>> clang: warning: t.hpp: 'linker' input unused when '-fsyntax-only' is
>>> present
>>> ddunbar at giles:tmp$ gcc -fsyntax-only t.hpp
>>> i686-apple-darwin10-gcc-4.2.1: t.hpp: linker input file unused because
>>> linking not done
>>> ddunbar at giles:tmp$
>>> --
>>>
>>> - Daniel
>>>
>>>> Nuno
>>>>
>>>> ----- Original Message -----
>>>> I don't buy it, I think something else is wrong. (My) gcc doesn't
>>>> recognize .hpp as c++ either, and this change is to clang-cc not the
>>>> driver, which always passes -x. So I don't get what this fixes?
>>>>
>>>> - Daniel
>>>>
>>>> On Tue, Nov 17, 2009 at 7:36 AM, Nuno Lopes <nunoplopes at sapo.pt> wrote:
>>>>>
>>>>> Author: nlopes
>>>>> Date: Tue Nov 17 09:36:35 2009
>>>>> New Revision: 89100
>>>>>
>>>>> URL: http://llvm.org/viewvc/llvm-project?rev=89100&view=rev
>>>>> Log:
>>>>> recognize .hpp as c++ (needed for ./configure'ing boost)
>>>>>
>>>>> Modified:
>>>>> cfe/trunk/lib/Frontend/FrontendOptions.cpp
>>>>>
>>>>> Modified: cfe/trunk/lib/Frontend/FrontendOptions.cpp
>>>>> URL:
>>>>>
>>>>> http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Frontend/FrontendOptions.cpp?rev=89100&r1=89099&r2=89100&view=diff
>>>>>
>>>>>
>>>>>
>>>>> ==============================================================================
>>>>> --- cfe/trunk/lib/Frontend/FrontendOptions.cpp (original)
>>>>> +++ cfe/trunk/lib/Frontend/FrontendOptions.cpp Tue Nov 17 09:36:35 
>>>>> 2009
>>>>> @@ -25,7 +25,7 @@
>>>>> .Case("mii", IK_PreprocessedObjCXX)
>>>>> .Case("C", IK_CXX)
>>>>> .Cases("C", "cc", "cp", IK_CXX)
>>>>> - .Cases("cpp", "CPP", "c++", "cxx", IK_CXX)
>>>>> + .Cases("cpp", "CPP", "c++", "cxx", "hpp", IK_CXX)
>>>>> .Case("cl", IK_OpenCL)
>>>>> .Default(IK_C);
>>>>> }