Re: Synopsis Question

I should note that these are changes I made to get things working under 
MSVC.  Other than adding the array information to Decoder.cc, I wasn't 
trying to add functionality.

Stefan Seefeld wrote:

> Luke Petre wrote:
>
>> I've attached the patch, it addresses the following:
>>  - Fixes Bugs in DefaultArgumentFinder
>>   + Uninitialized variables were passed for params and default_args
>>   + DefaultArgumentFinder was designed to operate on references to 
>> variables which were passed to the constructor.  Made that logic work 
>> correctly.
>
>
> Thanks for catching this. I obviously didn't test that code or else I 
> would
> have cought that stupid typo myself.
>
>> - MSVC Compilation fixes
>>   + Buffer.cc needs an include for algorithm
>>   + Lexer.hh needs an include for string
>>   + Path-win32.cc needs includes for direct.h (mkdir), and algorithm
>
>
> Fine. Please note that I'm going to phase out those platform-specific
> bits (such as the Path API) in favor of boost libs (such as 
> boost.filesystem).
> I'll apply the change, but please expect it to be relatively 
> short-lived :-)
>

That sounds great to me.

>>   + Named the anonymous namespace in ErrorHandler.cc
>
>
> Can you explain that change ?


MSVC would error that the symbol 'callback' was not defined in the 
global scope.  So I named the anonymous namespace _ErrorHandler in order 
to get access to that global variable.

>
>>   + Duplicated function params in swalker.hh
>
>
> see below.
>
>>   + type.cc, no such thing as Type::Type.  changed to Types::Type
>
>
> Funny that this error didn't get caught by the compiler !

see above.  the change in swalker.hh should really have been caught by 
the compiler too.

>
>>   + Display.hh, s.begin() returns a const_iterator
>
>
> Ok.
>
>> - Removed Python.h #include where possible
>>   + Unnecessary in linkstore.cc, filter.hh, swalker.cc
>>   + Forward declared PyObject in filter.hh
>>
>> - Added const friendly apis in type.hh and ast.hh
>> - Added Type::Array support to Decoder.cc
>
>
> I'm inclined not to apply any (non-critical) enhancements to the
> old AST translator (i.e. stuff in Synopsis/Parsers/Cxx) as I'm
> working hard to replace that code entirely. In fact, my plan is
> to make the next release as soon as the transition is done.
>
> In that light I would particularly welcome any help on the *new*
> code to bring it to the same level as the old translator so we
> don't get any regressions from the switch and only then add enhancements.

Where is the code for the new translator?  I'm sorry to say that I 
haven't been keeping up with the work you've been doing, but I'll be 
happy to try the new stuff out when I find some time.

>
> I believe the new code is much clearer and easier to understand,
> lowering the barrier for potential contributors. <wink/>
>
> Also, speaking of regressions, I would gladly accept patches
> that add new tests to the testing framework, so we have a better
> chance to catch regressions.
>
> Thanks again for your contribution !
>
My pleasure.

> Regards,
>         Stefan
>