Introducing POSIX-friendly type names

Torbjörn Granlund tg at
Mon May 30 12:42:40 CEST 2022

John Scott <jscott at> writes:

  GMP currently defines implementation-reserved identifiers all over the
  place, which is technically undefined behavior (this is how feature test
  macros like _POSIX_C_SOURCE work). In practice, this doesn't appear to
  hinder portability, but I'd like to focus on the number one way this
  affects applications: the type names.

  POSIX states that any system header is allowed to introduce a type
  ending in '_t', and as such GMP defining type names ending in '_t' can
  lead to name clashes. I wonder if, to accommodate application developers
  who want to write conforming POSIX applications, if GMP could introduce
  new type names without the _t suffix, perhaps by introducing new type
  names with the _t removed?

Sorry, this is a non-starter.

POSIX does not dictate things outside of POSIX.

Besides, I very much doubt that POSIX would add types which clash with
well-established libraries.

Please focus on real problems, not highly hypothetical ones.  Life is
difficiult as it is.

Please encrypt, key id 0xC8601622

More information about the gmp-discuss mailing list