[Pkg-pascal-devel] Castle Game Engine 6.0 release in Debian

[Michalis Kamburelis]

Paul Gevers wrote:
> That helped indeed. Now only "new" architectures are still failing. I
> had a shot at arm64, by applying this patch¹. 

Thank you, I applied your patch!

>However, after that it
> fails in the test-suite and it is all mystery for me². 

Hm, I don't understand it either for now... It crashes when calculating
some simple floating-point expressions using CastleScript, like "-10 * Pi".


> Also powerpc
> fails in the test-suite³. 

This is understandable: Our castlecompositeimage.pas is not correct on
big-endian machines. It should be fixed by
https://github.com/castle-engine/castle-engine/commit/8cc07193340a907e968d029c05a8759ba6b34b18

Please note that I fixed it without actually testing on a big-endian
machine. I added actually testing it (both by running the testsuite and
inspecting visually whether DDS images look sensibly) to my TODO list. I
should just make a virtual machine with such system.

>With ppc64, it looks like there are two CPU*
> definitions that trigger, so code in
> tools/build-tool/toolarchitectures.pas becomes invalid.

I see, CPUPOWERPC is for both CPUpowerpc32 and CPUpowerpc64 (
http://wiki.freepascal.org/Platform_defines ).

Fixed by trivial
https://github.com/castle-engine/castle-engine/commit/e02a4ceeb4b64f01c4a0f9b45a453a66cf6a6417
, changing

  {$ifdef CPUpowerpc} powerpc {$endif}

to

  {$ifdef CPUpowerpc32} powerpc {$endif}

(The name is still "powerpc", not "powerpc32", for compatibility. FPC's
FpMkUnit does it like this too.)

Regards,
Michalis