On 2016-08-27 00:24, Paul Eggert wrote:
> The main sticking point in this particular case is support for printf
> formats like %td. This requires library support, which in practice
> trails behind compiler support. We can reasonably ask people to use a
> C99ish compiler even when running on older platforms (just install GCC).
> It hasn't been so easy to insist on a C99ish C library.
*snip*
> PRI?PTR are not the right macros to use here. The values in question are
> of type ptrdiff_t, and ptrdiff_t != intptr_t on some (admittedly
> unusual) platforms.
Are you trying to argue that there exist some non-theoretical system,
where it's useful to actually run diffutils, and where both
intptr_t < long int and
intptr_t < ptrdiff_t
holds? (i.e. a system that would regress with the patch)
Because if there is no such system, then PRIdPTR w/ intptr_t will work
better than %ld w/ long int (pending the time when %td w/ ptrdiff_t is
generally available).
Cheers,
Peter
