On Sun, 2006-11-05 at 12:38 +0100, Sven Panne wrote: > Am Freitag, 3. November 2006 11:37 schrieb Ross Paterson: > > Fri Nov 3 02:35:59 PST 2006 Ross Paterson <[EMAIL PROTECTED]> > > * fill in Image routines in X11 (#982 from Frederik Eaton) > > > > M ./Graphics/X11/Types.hsc +12 > > M ./Graphics/X11/Xlib/Misc.hsc -11 +7 > > M ./Graphics/X11/Xlib/Types.hsc -1 +9 > > M ./X11.cabal -1 +2 > > [...] > > FYI: I've made this patch compile by telling the X11 headers to use normal > prototypes (not macros) for the image functions.
Good. Yeah, this issue of macros vs proper C functions is going to bite us more if/when ghc moves away from compiling via C by default. > In addition I have 3 remarks about the X11 package: > > * Strictly spoken, the majority of types are wrong: An Int is not a CInt, > and the binding will fail on platforms where the actual representation is > different. There should be a general code review of all the types in the > "foreign import"s. Yes, this will fail on most 64 bit machines where CInt is 32bit but Int is 64bit. > * The use of hsc2hs is not really needed, because the X11 prototypes and > #defines are fixed for all eternity because of ABI issues. This situation is > quite similar to OpenGL: There are different implementations, but they all > have to agree on a single ABI. Are there no structures with offsets etc? Those change between arches. Duncan
