[CMake] SWIG syntax errors

[Adam A Smith]

On Mon, 8 Jun 2009, Adam A Smith wrote:

> On Sun, 7 Jun 2009, Mathieu Malaterre wrote:
>
>> On Sat, Jun 6, 2009 at 2:19 AM, Adam A Smith<aasmith at cs.wisc.edu> wrote:
>>> I'm curious if someone has run into this before:
>>> 
>>> I'm using CMake to make some SWIG .i files.  I then use SWIG on those 
>>> files,
>>> and it tells me that I have a syntax error.  (I don't know SWIG syntax 
>>> that
>>> well, but I'm learning.)  I thought I'd ask all of you if this is 
>>> something
>>> you were familiar with.
>>> 
>>> Details:
>>> 
>>> There are several .cpp files in the C++ library I'm trying to translate to
>>> Perl, Python, etc.  (Doesn't matter what language I specify--the same 
>>> error
>>> comes up.)  For example, I might tell it:
>>> 
>>> $ swig -c++ -perl5 TextTools.cpp.i
>>> 
>>> and get this error:
>>> 
>>> TextTools.cpp.i:49: Error: Syntax error in input(1).
>>> 
>>> The .i file made from each .cpp file differs, of course.  But the syntax
>>> error is always here:
>>> 
>>> namespace std __attribute__ ((__visibility__ ("default"))) {
>>> # 245 "/usr/include/c++/4.3/i486-linux-gnu/bits/c++config.h" 3
>>> }
>>> 
>>> SWIG lists the first line (namespace std ...) as being the one with the
>>> error.  (So in my above example, that's line 49.)
>>> 
>>> Does anybody have an idea?
>> 
>> this is clearly a question for the swig mailing list, right ?
>> 
>> anyway I would add the following in my swig interface file:
>> 
>> #define  __attribute__(x)
>> 
>> 2cts
>> -- 
>> Mathieu
>> Ps: I am pretty sure this is not the root of the problem, you should
>> check the very begining of the errors reported...
>> 
>
> Thanks for the reply. :)
>
> Yes, I asked the SWIG mailing list as well.  But I was wondering if perhaps 
> it was a bug in the CMake output, since the code in the .i file doesn't 
> compile at all.  My main question (on this mailing list) wasn't so much "Can 
> somebody fix this?" as "Has anybody seen such a problem before?"
>
> Regardless, I sent the *complete* set of error messages.  There's just this 
> single syntax error.
>
> Thank you,
> Adam

Sorry--forgot to tell you the actual *results* of trying that.  It kind of 
worked.  It eliminated that error, but then told me of a new syntax error 
down the line.  This is a little confusing to me--since the .i file is all 
automatically generated.

Could you tell me what adding that #define line does?  Does it just tell 
it to expect a function/macro called __attribute__, and that the program 
shouldn't choke on it?

Thanks again,
Adam