Use of asprintf() is not POSIX-compliant about eudev HOT 11 CLOSED

I think trying to be POSIX-compliant on something that is requiring AF_NETLINK is a little crazy. If somebody wants to use this on non-Linux, they have to support AF_NETLINK anyway, so adding asprintf is not unreasonable to do at the same time.

from eudev.

uclibc and musl support both

from eudev.

Yeah I'm not sure how much of an idealist I am here. We can leave this open as a reminder, but asprintf is pretty meat and potatoes these days. Its not in posix and I don't know about the BSD world, but the latter doesn't concern us since eudev is Linux specific.

from eudev.

Oops, wrong button!

from eudev.

clang -D_BSD_SOURCE=1 might solve this problem.

from eudev.

You could consider providing your own asprintf implementation to stay compatible with the standard.

from eudev.

Is this still a concern? I don't see those warnings while compiling.

Adding #define _GNU_SOURCE is reasonable, though, since variadic functions with implicit declarations can lead to weird issues.

from eudev.

@ericonr Did you compile on musl? Its not an issue on glibc, but I recall that it was an issue on musl.

from eudev.

Yes, on musl.

That said, I can't find out where, if anywhere, _GNU_SOURCE is being passed to it... Or if we're just getting lucky with defaults. I will see if I can come up with a clean fix to be sure the issue never happens.

from eudev.

I recently explored this idea with another project. While asprintf() may not be part of POSIX, it really is virtually everywhere - muslc, glibc, FreeBSD, and NetBSD Not sure, but it might even be available through MINIX.

from eudev.

The makefiles for files that use this function include this snippet, and config.h defines _GNU_SOURCE

	-include $(top_builddir)/config.h

So I think this can be closed (even if it's an ugly solution).

from eudev.