https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122638
--- Comment #6 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> --- > --- Comment #5 from Eric Gallager <egallager at gcc dot gnu.org> --- > (In reply to [email protected] from comment #2) >> macOS is a different beast, though: 10.13 and 12 both have ancient 4.8 >> bundled, while 26 apparently doesn't bundle makeinfo any longer. Given >> that, I'd suggest 5.0 is fine since one has to build your own anyway. >> It's Iain's call in the end. >> >> IIRC the makeinfo step is skipped if there's no makeinfo present. Maybe >> we should check for a matching version somewhere instead of relying on >> install.texi alone? >> >> So for me, 5.0 would be perfectly fine and quite reasonable. > > Not everyone is on 26 yet, though, so some people might still be on machines > that have the shipped 4.8 on them. I think bumping the requirement from 4.7 to > 4.8 would be as far as I'd be comfortable going for the next release. However, such a requirement and merely documenting it in install.texi is meaningless unless it's enforced with a version check and makeinfo treated as missing if too old *and* it's regularly tested that make info actually works with 4.8. I certainly won't be doing the latter: as I found when trying, there are already too many places where it fails right now. While this can probably be fixed, someone has to do the work.
