|
From: | Philip Prindeville |
Subject: | Re: ac_cv_sizeof_X, et al. |
Date: | Mon, 14 Jul 2008 23:22:08 -0700 |
User-agent: | Thunderbird 2.0.0.14 (Windows/20080421) |
Bob Friesenhahn wrote:
On Mon, 14 Jul 2008, Philip Prindeville wrote:And is there a "autoconf-lint" to check for using broken names, like "ac_cv_sizeof_long_long_unsigned_int" instead of whatever is preferred?I prefer ac_cv_sizeof_long_long_unsigned_int because it is my nature to be pedantic and it sounds bigger. Is that bad?
Well, it kind of begs the question: if ac_sizeof_long_long_int != ac_sizeof_long_long_unsigned_int, then what happens when you assign a LONGLONG_MAX to a signed long long, and then copy that into an unsigned long long, or vice versa...
These same tests can test for any type from a header file. It is pretty difficult to "optimize" them.
Well, I think there is a pretty common core set that takes care of 99% of all cross compilation:
char short int long long long double char/void pointer size_t offset_t ptrdiff_t time_t
If <stdint.h> and <inttypes.h> were available everywhere then much of this gobbly gook could go away. Maybe in another 10 years or so that will be the case.
Not as long as some ANSI or ISO group is in charge of it. IEEE almost got it right, but not quite. Then again, not everyone is POSIX, either.
-Philip
Bob
[Prev in Thread] | Current Thread | [Next in Thread] |