On Wed, Oct 09, 2024 at 05:36:42PM +0200, David Sterba wrote: > On Wed, Oct 09, 2024 at 05:45:10PM +0300, Andy Shevchenko wrote: > > On Tue, Oct 08, 2024 at 06:10:32PM +0200, David Sterba wrote: > > > On Mon, Oct 07, 2024 at 06:16:10PM -0500, Ira Weiny wrote: ... > > > > +static inline bool range_overlaps(struct range *r1, struct range *r2) > > > > > > I've noticed only now, you can constify the arguments, but this applise > > > to other range_* functions so that can be done later in one go. > > > > Frankly you may add the same to each new API being added to the file and > > the "one go" will never happen. > > Yeah, but it's a minor issue for a 28 patchset, I don't know if there > are some other major things still to do so that a v5 is expected. At least seems printf() changes have to be amended, so I think v5 is warranted anyway. > If anybody is interested, reviewing APIs and interfaces with focus on > some data structure and const is relatively easy, compile test is > typically enough. Except the cases where a const pointer has to be passed thru non-const (or integer) field in a data structure. Tons of the existing examples is ID tables that wanted to have kernel_ulong_t instead of const void * in driver data field. > The hard part is to find the missing ones. There's no > compiler aid thad I'd know of (-Wsuggest-attribute=const is not for > parameters), so it's been reading a file top-down for me. Yeah... > > So, I support your first part with > > constifying, but I think it would be rather done now to start that "one > > go" to happen. > > Agreed, one patch on top is probably the least intrusive way. -- With Best Regards, Andy Shevchenko
