On Tue, Apr 25, 2023 at 08:24:01AM +0300, Matti Vaittinen wrote: > On 4/24/23 18:22, Andy Shevchenko wrote: > > On Mon, Apr 24, 2023 at 04:10:09PM +0300, Matti Vaittinen wrote: ... > > This... > > > > > +#include <linux/bits.h> > > > > ...is guaranteed to be included by this. > > > > > +#include <linux/bitops.h> > > Out of the curiosity - do we have a rule and rationale for explicitly > including headers with 'stuff' we use Vs. trusting some header being > included by another one? I've not thought much of this so I don't know if > there are any pros/cons? That's what we are starving for actually. Currently this is a tribe knowledge which one gets while being involved into Linux kernel development for a long time and being capable of keeping an eye on tree wide, library or similar changes. I would love to see some (preferably generated) list of the header dependencies. Yet, the header dependency hell should be solved meanwhile (see Ingo's 2k+ patch series). -- With Best Regards, Andy Shevchenko
