[Qgis-developer] QgsRasterBlock.setColor(size_t, QRgb) in SIP

[Martin Dobias]

On Thu, Oct 24, 2013 at 9:40 PM, Radim Blazek <radim.blazek at gmail.com> wrote:
>
> Then I get
>
>>>> block.setColor( 0, QColor ().rgb() )
> Traceback (most recent call last):
>   File "<input>", line 1, in <module>
> TypeError: arguments did not match any overloaded call:
>   QgsRasterBlock.setColor(int, int, int): not enough arguments
>   QgsRasterBlock.setColor(int, int): argument 1 has unexpected type 'int'
>
> works with "unsigned long". In C++ I have size_t defined as "long
> unsigned int" (gives compilation error if used in sip file) which
> should be the same as "unsigned long" I believe.
>
> The problem is wrong definition of size_t in sip file. It should not
> work for you as well, right? If you are on Linux you get "typedef
> unsigned int size_t" in sip which is not the same as size_t in C++. I
> guess that it accidentally works because by chance the memory used by
> wider type is zeroed.

Yes I also think it works just accidentally.

>
>> There was a thread on size_t in PyQt some time ago:
>> http://www.riverbankcomputing.com/pipermail/pyqt/2006-March/012753.html
>
> Yes, I saw, quite old and no solution. If I look into stddef.h
> (/usr/lib/gcc/x86_64-linux-gnu/4.6/include/stddef.h) the typedef of
> size_t is quite complex.
>
> So how can we reliably define size_t in sip file?

We should probably try to compile and run a tiny program (during CMake
configure phase) that will return the value of sizeof(size_t) and from
that generate a helper .sip file with correct definition for size_t.
Not extremely elegant, but should work. The problem will be again with
cross-compilation (e.g. Android) as we cannot run such utility program
on target CPU, but we can hardcode the definition for such platforms.
But maybe there's a better way I am not aware of...


> I mean what is the best practice to do that in QGIS. Would not it be
> better to have shared definitions separated into some common.sip?

Yes, that would be better.


Martin