On Mon, Feb 07, 2022 at 03:32:10PM -0600, Eric Sandeen wrote: > On 2/7/22 11:09 AM, Darrick J. Wong wrote: > > On Mon, Feb 07, 2022 at 12:05:41PM +1100, Dave Chinner wrote: > >> On Fri, Feb 04, 2022 at 04:36:18PM -0800, Darrick J. Wong wrote: > >>> On Fri, Feb 04, 2022 at 05:18:12PM -0600, Eric Sandeen wrote: > > ... > > >>>> Do we /need/ to build fully functional xfsprogs on old userspace? > >>>> (really: systems with old kernel headers?) How far back do we go, > >>>> I wonder? Anyway... > >>> > >>> TBH we could probably get rid of these entirely, assuming nobody is > >>> building xfsprogs with old kernel headers for a system with a newer > >>> kernel? > >> > >> Just fiddle the autoconf rules to refuse to build if the system > >> headers we need aren't present. It just means that build systems > >> need to have the userspace they intend to target installed in the > >> build environment. > > > > GETFSMAP premiered in 4.12, so I'm going to take this response (and the > > lack of any others) as a sign that I can respin this patch to require > > recent kernel headers instead of providing our own copy. > > Sounds reasonable, thanks. Maybe in the future when we add stuff like > this for bleeding edge interfaces we can mark the date, and mark another > one in what, a year or two, as a reminder for removal. That's what we've done in the past - provide our own copy until the system headers catch up and then remove our copy. Note that this may cause angst with lesser used C libraries (like musl), but they need to keep up with new kernel APIs to be really useful, anyway... Cheers, Dave. -- Dave Chinner david@xxxxxxxxxxxxx
