Use of %td in printf

Use of %td in printf

[Eli Zaretskii]

Commit c136f93d causes us use %td in printf formats, but is that
feature portable enough?  (Admittedly, the code which uses that is
usually not compiled in, but still.)

Re: Use of %td in printf

[Paul Eggert]

Eli Zaretskii wrote:
> Commit c136f93d causes us use %td in printf formats, but is that
> feature portable enough?

Yes. %td is in C99 and Emacs has been using the feature for some time (Emacs 26 
uses it).

Re: Use of %td in printf

[Eli Zaretskii]

> Cc: emacs-devel@gnu.org
> From: Paul Eggert <eggert@cs.ucla.edu>
> Date: Wed, 3 Jul 2019 00:33:42 -0700
> 
> Eli Zaretskii wrote:
> > Commit c136f93d causes us use %td in printf formats, but is that
> > feature portable enough?
> 
> Yes. %td is in C99 and Emacs has been using the feature for some time (Emacs 26 
> uses it).

AFAIR, we require a C99 compiler, but not a C99 C library.

Re: Use of %td in printf

[Paul Eggert]

Eli Zaretskii wrote:
> AFAIR, we require a C99 compiler, but not a C99 C library.

That was a while ago. These days it's safe to assume %td and Emacs has been 
doing so for some time without any problems.

I should mention that the situation is more complicated than "we require a C99 
compiler". Emacs does not require full support for C99, either in the compiler 
or the library. (If it did, we couldn't use Microsoft's C compiler, which does 
not support all of C99.) Emacs does assume many C99 features, though, and %td 
has been one of these features for a while.

While we're on the topic, one of these days I was thinking of removing 
printmax_t and related types and macros from src/lisp.h, as %jd is considerably 
simpler and should be universal by now. Even when I added that stuff back in 
2011 the concern about portability to old C libraries lacking %jd support was 
mostly theoretical; I think the last holdout was Solaris 8, which Oracle stopped 
supporting in 2012. Nowadays those pre-C99 libraries are museum pieces and we 
can drop the printmax_t hacks.

Re: Use of %td in printf

[Eli Zaretskii]

> Cc: emacs-devel@gnu.org
> From: Paul Eggert <eggert@cs.ucla.edu>
> Date: Wed, 3 Jul 2019 01:04:33 -0700
> 
> Eli Zaretskii wrote:
> > AFAIR, we require a C99 compiler, but not a C99 C library.
> 
> That was a while ago. These days it's safe to assume %td and Emacs has been 
> doing so for some time without any problems.

Thanks, I just wanted to be sure.  I don't really know of any platform
where this could cause problems.

> I should mention that the situation is more complicated than "we require a C99 
> compiler". Emacs does not require full support for C99, either in the compiler 
> or the library. (If it did, we couldn't use Microsoft's C compiler, which does 
> not support all of C99.)

We don't support building with the Microsoft compiler for quite some
time now.  The MS-Windows port is built using GCC, not MSVC.

> While we're on the topic, one of these days I was thinking of removing 
> printmax_t and related types and macros from src/lisp.h, as %jd is considerably 
> simpler and should be universal by now. Even when I added that stuff back in 
> 2011 the concern about portability to old C libraries lacking %jd support was 
> mostly theoretical; I think the last holdout was Solaris 8, which Oracle stopped 
> supporting in 2012. Nowadays those pre-C99 libraries are museum pieces and we 
> can drop the printmax_t hacks.

If all the platforms we care about support that, fine.