On Mon, Feb 16, 2009 at 1:26 PM, Chris Robinson <chris.kcat@gmail.com> wrote:
On Monday 16 February 2009 9:38:19 am Seth Shelnutt wrote:
> I had an interesting thought the other day, and that is to having some
> built in support for forwarding windows dlls to linux .so's.

IIRC, this kind of thing is generally discouraged, except in cases where
needed (eg. opengl32). Part of the problem is internal differences.. for
example, what would a Linux .so do if it's given a Win32 filename path? Other
problems would be if the Linux equivalent wants a Window for a function
argument and the DLL wants an HWND instead, or if the DLL has more/different
functions (eg. CUDA on Win32 has functions for dealing with D3D objects; CUDA
on Linux doesn't, and it wouldn't be straight forward to even implement them
through wined3d).

Something like OpenGL, and even OpenAL to some degree, would directly benefit
from having the DLLs forwarded to their host equivalents as it provides better
access to the hardware and better integration with the host system. Something
like zlib or ogg/vorbis and stuff wouldn't though, since the code should
largely be the same, and save for some bugs/inefficiencies in Wine (which
should be fixed), would work identically.



Does Wine yet have the capability to interface with HAL for Win32 hardware access similar to NT? It looks like it doesn't from all this talk of forwarding DLLs. What we should do instead of trying to forward DLLs, which is asking for more trouble than its worth, is try to get the NT layer to connect to UNIX HAL so that DLLs can link directly to HAL and operate the hardware.