Re: Android port

[Po Lu]

Eli Zaretskii <eliz@gnu.org> writes:

> That doesn't answer my question, AFAICT.  In the current master we
> have no uses of asprintf and vasnprintf, so I asked whether it is
> needed on the branch, and if so, whether the MS-Windows build uses the
> code where these two functions are called.

It doesn't, but vfprintf-posix actually replaces all the printf
functions, not just (vasn)printf.  The replacement functions are not
necessary on Windows, however.

> The way to override Gnulib tests that conclude that some libc function
> should be replaced is not to override the feature test (unless that
> feature is supported, but Gnulib doesn't know about it -- which can
> only happen if we implement the library function inside Emacs).  The
> way to override those is to exclude the relevant Gnulib modules from
> the Windows build via nt/gnulib-cfg.mk.

There is already:

OMIT_GNULIB_MODULE_vasnprintf = true
OMIT_GNULIB_MODULE_vasprintf = true
OMIT_GNULIB_MODULE_vfprintf-posix = true

within gnulib-cfg.mk, but fudging with the feature tests is also
necessary to generate the Gnulib stdio.h header correctly; absent that,
it tries to provide definitions for its printf replacements, which does
not work.

IIRC manipulating the feature tests in such a fashion was proposed by
one of the Gnulib developers when the W32 build was last tested in
March, but my memory of that is indistinct and I might've read this
elsewhere.