Re: [GIT PULL] tracing: Fixes and clean ups for v5.14

[Randy Dunlap]

On 8/12/21 11:57 AM, Randy Dunlap wrote:
> On 8/12/21 11:17 AM, Linus Torvalds wrote:
> > On Thu, Aug 12, 2021 at 8:04 AM Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
> > > I just used some default settings. I didn't choose to use -Wmain.
> >
> > What broken distro, what broken gcc version?
>
> openSUSE 15.3
> gcc (SUSE Linux) 7.5.0
>
> > We can certainly add a -Wno-main for this case. We already do that for
> > a lot of other idiotic warnings like -Wno-pointer-sign.
>
> That's what my first patch did, but Steven didn't like it.
>
> > But when we do so, I want the exact tool and distro version named and
> > shamed. Because I sure don't see that warning, and from what I can
> > tell, most other people don't see it either.
> >
> > So it's almost certainly your distro that has configured the gcc
> > install incorrectly - or some new gcc version that makes new insane
> > defaults. The commit message should talk about those kinds of details,
> > exactly so that people like me get an explanation for why we'd need
> > that odd '-Wno-main' flag.
> >
> > Maybe even the line in the Makefile should have it. Like that
> > -Wno-pointer-sign thing does:
> >
> >    # disable pointer signed / unsigned warnings in gcc 4.0
> >    KBUILD_CFLAGS += -Wno-pointer-sign
> >
> > just because unexplained random compiler flags are a bad thing (the
> > same way unexplained random code changes due to them are bad)

I probably misread that a bit. To be clear, my patch only set -Wno-main
for this one source file, not for the entire kernel tree, whereas Steven
wanted it to be set to the full kernel tree.


--
~Randy