[CMake] Gyp VS CMake

[Nicolas Desprès]

On Mon, Feb 21, 2011 at 6:38 PM, Bill Hoffman <bill.hoffman at kitware.com> wrote:
> On 2/21/2011 12:16 PM, j s wrote:
>
>>
>>    LD_PRELOAD isn't that bad of a hack, actually. Intercepting open(),
>>    read(), and possibly mmap() should cover most cases.
>>
>>
>> So
>> cl.exe /showincludes
>>
>> doesn't work?  According to this, it is available.
>>
>> http://www.conifersystems.com/2008/10/09/dependencies-from-showincludes/
>>
>>
>
> Although slower that what we have.  I did write a depend system once that
> used cl.exe.  It basically ran the cpp over the code (cl -E) , and grepped
> out all the #line information.  It has the advantage of not using an system
> tricks, and all of the compilers we support can be made to print out #line
> numbers with a cpp output.  We don't have any free cycles to work on this at
> Kitware right now.
>

Just an idea, but wouldn't it be possible to make the dependency
scanner used by CMake customizable.  I mean one may want to use its
own dependency scanner, instead of the built-in one, for whatever
reason.  I see at least two use cases:

1/ To work around #if 0 bug in cmake dependency scanner
http://www.vtk.org/Wiki/CMake_FAQ#CMake_dependency_scanner

2/ Some project generate code and may have a higher level point of
view regarding files dependencies and could generate dependency files
faster.

This may also helps to add support for build-system like tup in CMake.

-- 
Nicolas Desprès