On 2010-5-2 4:20 PM, Igor Kovalenko wrote:
[...] This reminds me to ask a question if we have to encode all properties as 32bit integers.
By tradition, yes - the assumption is that you write bindings for all implementations, not just for your particular implementation, so you can't assume the implementation has 64-bit cells. Also, if you look at the 1275 standard for "encode-int" (page 140 or so), it states specifically that it produces a string of /four/ bytes, with high-order byte being first in order (i.e., big-endian ordering).
At the moment there are a few places which encode cells (64bit on sparc64) to properties. F.e. it breaks linux kernel code which expects framebuffer properties to be 32bit whereas init_video will encode 64bit values.
If they are encoding ints (with encode-int), they *must* be 32-bit items. If they are encoding byte-arrays, they can do anything.