Support both VaList::arg() and VaList::next_arg() nightly API

[renkenono]

VaList::arg() appears to be renamed to VaList::next_arg() in the current nightly which is 1.98.0-nightly (f28ac764c 2026-06-23) as of time of checking. Adding support for the latter would enable projects requiring such recent nightly versions to compile printf-compat with no issues.

I haven't done much research to see how we can determine the availability of methods at compile-time and/or perhaps merely introduce a feature flag to opt-in into the future next_arg API instead but any approach would be fine by me!

Again, thanks for the crate!

[renkenono]

I wonder if an opt-in feature to support the older API is a better approach.

#[cfg(not(feature = "legacy-va-list"))]
trait NextArgCompat {
    unsafe fn next_arg<T: VaArgSafe>(&mut self) -> T;
}

#[cfg(not(feature = "legacy-va-list"))]
impl NextArgCompat for VaList<'_> {
    #[inline]
    unsafe fn next_arg<T: VaArgSafe>(&mut self) -> T {
        self.arg()
    }
}

[nicholasbishop]

VaList::arg() appears to be renamed to VaList::next_arg() in the current nightly which is 1.98.0-nightly (f28ac764c 2026-06-23) as of time of checking. Adding support for the latter would enable projects requiring such recent nightly versions to compile printf-compat with no issues.

I'm a little confused by the phrasing here -- it sounds like you're saying that current nightly isn't supported because we don't use next_arg, but we did make that change in c8cdff5, and the latest 0.4.0 release contains that.

[renkenono]

Apologies, I mixed up the API naming as I was writing this issue, the idea is to support older nightly compilers which do not offer the next_arg API.

[nicholasbishop]

Ah OK. Can you say more about why you want that? You should be able to use an older release (I think 0.3.1 should work) if you are pinned to an older nightly.

[renkenono]

It appears that the toolchain I use, rustc 1.96.0-nightly (38c0de8dc 2026-02-28), has both the VaList lifetime changes as well as VaList::next_arg() rename which makes this issue useless, I'm sorry for the unnecessary issue. Thanks for your help!