On Tue, 03 Oct 2023, Jonathan Corbet <corbet@xxxxxxx> wrote: > "Matthew Wilcox (Oracle)" <willy@xxxxxxxxxxxxx> writes: > >> kernel-doc checks were initially enabled only for builds which had extra >> warnings enabled. We have now eliminated enough kernel-doc warnings that >> we can enable kernel-doc checking by default. This comes at a slight >> cost; for an allmodconfig build, make -j8 fs/ timings on my laptop >> increase by less than 5%: >> >> before real 4m7.456s 4m4.416s 4m6.663s >> after real 4m18.960s 4m21.566s 4m23.234s >> before user 29m35.370s 29m11.036s 29m30.092s >> after user 30m55.602s 31m10.918s 31m20.311s >> before sys 2m8.230s 2m6.392s 2m9.727s >> after sys 2m19.896 2m23.422s 2m25.762s >> >> This feels like a reasonable price to pay to force people to keep >> documentation up to date. >> >> Signed-off-by: Matthew Wilcox (Oracle) <willy@xxxxxxxxxxxxx> >> Reviewed-by: Carlos Bilbao <carlos.bilbao@xxxxxxx> >> --- >> scripts/Makefile.build | 4 +--- >> 1 file changed, 1 insertion(+), 3 deletions(-) > > So I finally got around to actually giving this a try ... > > It adds 1,095 warnings to an allmodconfig build here. If I merge that, > I suspect that next thing that happens will be One Of Those Emails from > Linus... and perhaps from others as well. > > If we had a series that drove the number to zero prior to this change, > it would be a different story. I'm kind of thrashing and don't think I > can do that in the near future, as nice as it would be. I suspect > there's not a lot of other folks just waiting for a chance to do this > either. > > As nice as it would be to have this, I don't think it would survive to a > mainline release if I tried to push it now. But maybe others disagree? I'll leave that decision up to the people who have to face the music. But an alternative would be to add more warning levels to kernel-doc, put the most verbose messages from the allmodconfig run behind some -Wall knob, always run kernel-doc without that, and make W=1 run with -Wall. It would be an iterative path forward. Now Someone(tm) just needs to implement that. ;) BR, Jani. -- Jani Nikula, Intel
