Re: [PATCH net-next v2 1/2] Makefile.extrawarn: Add symbol for W=1 warnings for today

From: Arnd Bergmann <arnd@arndb.de>
To: Andrew Lunn <andrew@lunn.ch>
Cc: Nick Desaulniers <ndesaulniers@google.com>,
	netdev <netdev@vger.kernel.org>,
	David Miller <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>,
	Masahiro Yamada <masahiroy@kernel.org>,
	Michal Marek <michal.lkml@markovi.net>,
	Rohit Maheshwari <rohitm@chelsio.com>,
	Linux Kbuild mailing list <linux-kbuild@vger.kernel.org>,
	clang-built-linux <clang-built-linux@googlegroups.com>
Subject: Re: [PATCH net-next v2 1/2] Makefile.extrawarn: Add symbol for W=1 warnings for today
Date: Mon, 5 Oct 2020 22:03:47 +0200	[thread overview]
Message-ID: <CAK8P3a1qS8kaXNqAtqMKpWGx05DHVHMYwKBD_j-Zs+DHbL5CNw@mail.gmail.com> (raw)
In-Reply-To: <20201005194913.GC56634@lunn.ch>

On Mon, Oct 5, 2020 at 9:49 PM Andrew Lunn <andrew@lunn.ch> wrote:
>
> > Sorry, to be more specific about my concern; I like the idea of
> > exporting the W=* flags, then selectively applying them via
> > subdir-ccflags-y.  I don't like the idea of supporting W=1 as defined
> > at a precise point in time via multiple date specific symbols.  If
> > someone adds something to W=1, then they should need to ensure subdirs
> > build warning-free, so I don't think you need to "snapshot" W=1 based
> > on what it looked like on 20200930.
>
> That then contradicts what Masahiro Yamada said to the first version i
> posted:
>
> https://www.spinics.net/lists/netdev/msg685284.html
> > With this patch series applied, where should we add -Wfoo-bar?
> > Adding it to W=1 would emit warnings under drivers/net/ since W=1 is
> > now the default for the net subsystem.
>
> The idea with the date stamps was to allow new warnings to be added to
> W=1 without them immediately causing warnings on normal builds. You
> are saying that whoever adds a new warning to W=1 needs to cleanup the
> tree which is already W=1 clean? That might have the side effect that
> no more warnings are added to W=1 :-(

It depends a lot on what portion of the kernel gets enabled for W=1.

As long as it's only drivers that are actively maintained, and they
make up a fairly small portion of all code, it should not be a problem
to find someone to fix useful warnings.

The only reason to add a flag to W=1 would be that the bugs it reports
are important enough to look at the false positives and address
those as well. Whoever decided to enable W=1 by default for their
subsystem should then also be interested in adding the new warnings.

If I wanted to add a new flag to W=1 and this introduces output
for allmodconfig, I would start by mailing that output to the
respective maintainers.

        Arnd