|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |
To: Tom Yu <tlyu@MIT.EDU> Cc: krbdev@MIT.EDU From: Assar Westerlund <assar@sics.se> Date: 22 Oct 1997 00:02:45 +0200 In-Reply-To: Tom Yu's message of "Fri, 17 Oct 1997 23:44:52 -0400 (EDT)" Tom Yu <tlyu@MIT.EDU> writes: > So what do people think about getting rid of sized types completely? > In particular, krb5_int32 will always be a long, and krb5_int16 will > always be an int. Better yet, what about just using long and int > where we mean types at least 32 and 16 bits wide, accordingly? I > don't think that's inherently a bad thing. It will likely break some > binary compatibility, but considering Ken Hornstein's experiences with > with Crays and other associated confusion, I am strongly considering > it. Short answer: no Long answer follows... The only types we can count on being there are `at least n bits'. And of course, the code cannot depend on these having exactly n bits of precision or occupying exactly n bits in memory. And I think there are two advantages to calling them `foo_16' and `foo_32' instead of `int' and `long'. First, that's it's more obvious that we need a type of at least `n' bits. And second because it gives flexibility later to change things on strange machines. In C9X there's this nice header called <inttypes.h> which contains these types (for canonical values of 17): int17_t, uint17_t (these may or may not exist) int_least17_t, uint_least17_t (must exist) int_fast17_t, uint_fast17_t (must exist) and then (in the cases where there will be any difference) picking the *fast* makes the most sense to me. Of course, having foo_16 == int and foo_32 == long as a fall-back is the right thing. The way we have done things in Heimdal is mostly like above except that we (for historical reasons) call the types `int17_t' and `u_int17_t'. If these are defined when compiling we use these definitions, otherwise we generate a small program to calculate the precision of each integer type and generate a header from that. /assar
|
|
|
|
|
|
|
|
|
|
|
|
| home | help | back | first | fref | pref | prev | next | nref | lref | last | post |