APPROVE COMMIT 21.5
I've collapsed these into a single patch and pushed it.
Stephen J. Turnbull writes:
As usual, I'll give this a few days, then commit the lot.
Inspired by Jeff's Xkb patch, I cleaned up a bunch of warnings emitted
by clang (3.2--3.4beta versions) when XEmacs is built with
error-checkingx. With Jeff's patch, I now have zero warnings.
A lot of them are related to Jerry's recent (?) work on bignums (the
introduction of make_integer and friends). So Jerry, I'd appreciate
your opinion (some of them are a little ugly; mostly they appeal to me
as a very type-conscious hacker, but I'm not the one who maintains
that code, and the warnings related to make_integer and friends occur
because of bounds checks only in error-checking builds IIUC).
A generic comment: I believe we should avoid use of unsigned types,
almost at any cost. Careful analysis showed that the only use we have
for them is when external APIs use them. Then we should treat them
like legacy encodings, and convert them immediately to a Lisp type.
I left the unsigned integers in the code that deals with struct stat.
I don't understand the underlying APIs well enough to change it.
XEmacs-Patches mailing list