Re: [PATCH net-next v1 2/2] net: core: Sort headers alphabetically

[Alexei Starovoitov]

On Tue, Sep 12, 2023 at 10:05 AM Andy Shevchenko
<andriy.shevchenko@xxxxxxxxxxxxxxx> wrote:
>
> On Tue, Sep 12, 2023 at 06:53:23PM +0200, Paolo Abeni wrote:
> > On Tue, 2023-09-12 at 19:35 +0300, Andy Shevchenko wrote:
> > > On Tue, Sep 12, 2023 at 05:20:31PM +0200, Simon Horman wrote:
> > > > On Mon, Sep 11, 2023 at 06:45:34PM +0300, Andy Shevchenko wrote:
> > > > > It's rather a gigantic list of heards that is very hard to follow.
> > > > > Sorting helps to see what's already included and what's not.
> > > > > It improves a maintainability in a long term.
> > > > >
> > > > > Signed-off-by: Andy Shevchenko <andriy.shevchenko@xxxxxxxxxxxxxxx>
> > > >
> > > > Hi Andy,
> > > >
> > > > At the risk of bike shedding, the sort function of Vim, when operating
> > > > with the C locale, gives a slightly different order, as experssed by
> > > > this incremental diff.
> > > >
> > > > I have no objections to your oder, but I'm slightly curious as
> > > > to how it came about.
> > >
> > > !sort which is external command.
> > >
> > > $ locale -k LC_COLLATE
> > > collate-nrules=4
> > > collate-rulesets=""
> > > collate-symb-hash-sizemb=1303
> > > collate-codeset="UTF-8"
> >
> > I'm unsure this change is worthy. It will make any later fix touching
> > the header list more difficult to backport, and I don't see a great
> > direct advantage.
>
> As Rasmus put it here
> https://lore.kernel.org/lkml/5eca0ab5-84be-2d8f-e0b3-c9fdfa961826@xxxxxxxxxxxxxxxxxx/
> In short term you can argue that it's not beneficial, but in long term it's given
> less conflicts.

I agree with Paolo.

This is just code churn.
The includes will become unsorted eventually.
Headers might get renamed, split, etc.
Keeping things sorted is a headache.