Re: [PATCH RFC/RFT 0/2] W=1 by default for Ethernet PHY subsystem

[Jason Gunthorpe <jgg@nvidia.com>]

On Wed, Feb 10, 2021 at 10:43:29AM -0800, Jakub Kicinski wrote:
> On Wed, 10 Feb 2021 14:39:17 -0400 Jason Gunthorpe wrote:
> > On Sun, Sep 20, 2020 at 04:53:51PM +0200, Andrew Lunn wrote:
> > 
> > > How often are new W=1 flags added? My patch exported
> > > KBUILD_CFLAGS_WARN1. How about instead we export
> > > KBUILD_CFLAGS_WARN1_20200920. A subsystem can then sign up to being
> > > W=1 clean as for the 20200920 definition of W=1.  
> > 
> > I think this is a reasonable idea.
> > 
> > I'm hitting exactly the issue this series is trying to solve, Lee
> > invested a lot of effort to make drivers/infiniband/ W=1 clean, but
> > as maintainer I can't sustain this since there is no easy way to have
> > a warning free compile and get all extra warnings.  Also all my
> > submitters are not running with W=1
> > 
> > I need kbuild to get everyone on the same page to be able to sustain
> > the warning clean up. We've already had a regression and it has only
> > been a few weeks :(
> 
> Do you use patchwork? A little bit of automation is all you need,
> really. kbuild bot is too slow, anyway.

Yes, I bookmarked your automation scripts, but getting it all setup
for the first time seems like a big mountain to climb..

I already do compile almost every patch, the problem is there is no
easy way to get W=1 set for just a subset of files and still compile
each file only once.

If this makefile change was done then my submitters would see the
warning even before posting a patch. It is already quite rare I see
patches with compile warnings by the time they get to me. Even
unlikable checkpatch warnings have been fairly low..

Jason