Guy Harris wrote:
On Aug 16, 2009, at 9:32 PM, Kovarththanan Rajaratnam wrote:
> I'm not sure that casting a pointer to g_malloc() to a pointer to a
function taking a size_t as an argument would work on platforms where
sizeof(size_t) != sizeof(gulong); Win64 is, as far as I know, one such
platform, as it's LLP64, so "gulong" is 32 bits and "size_t" is 64
bits, and we do support Win64.
I really wish the GLib people had used gsize in a bunch of their APIs;
gulong and size_t are the same size in ILP32 and LP64 platforms, so
the only place where fixing that would cause problems would be LLP64
platforms such as Win64. Whether they *will* fix it is another
matter....
I would vote either for using Martin's patch or for writing a wrapper
around g_malloc() that takes a size_t as an argument.
I actually added the cast because I thought it would be safe for the
same reasons you've already mentioned: gsize is only an issue on Win64
and on this platform we bundle the newest version of GLib where the
gsize/gulong mismatch has already been fixed:
http://git.gnome.org/cgit/glib/commit/?id=1fcaf2fe8d5118451e5d9da3347f711b623d504d
Is it ok to keep r29451?
Best regards,
Kovarththanan Rajaratnam