Re: Linux 6.3-rc3

[Guenter Roeck]

On Mon, Mar 20, 2023 at 11:26:17AM -0700, Linus Torvalds wrote:
> On Mon, Mar 20, 2023 at 11:05 AM Nathan Chancellor <nathan@xxxxxxxxxx> wrote:
> >
> > On the clang front, I am still seeing the following warning turned error
> > for arm64 allmodconfig at least:
> >
> >   drivers/gpu/host1x/dev.c:520:6: error: variable 'syncpt_irq' is uninitialized when used here [-Werror,-Wuninitialized]
> >           if (syncpt_irq < 0)
> >               ^~~~~~~~~~
> 
> Hmm. I do my arm64 allmodconfig builds with gcc, and I'm surprised
> that gcc doesn't warn about this.
> 
> That syncpt_irq thing isn't written to anywhere, so that's pretty egregious.
> 
> We use -Wno-maybe-uninitialized because gcc gets it so wrong, but
> that's different from the "-Wuninitialized" thing (without the
> "maybe").
> 
> I've seen gcc mess this up when there is one single assignment,
> because then the SSA format makes it *so* easy to just use that
> assignment out-of-order (or unconditionally), but this case looks
> unusually clear-cut.
> 
> So the fact that gcc doesn't warn about it is outright odd.
> 
> > If that does not come to you through other means before -rc4, could you
> > just apply it directly so that I can stop applying it to our CI? :)
> 
> Bah. I took it now, there's no excuse for that thing.
> 
> Do we have any gcc people around that could explain why gcc failed so
> miserably at this trivial case?
> 

I have noticed that gcc doesn't always warn about uninitialized variables
in most architectures. The conditional btrfs build failure (only seen on
sparc and parisc) is similar: gcc is silent even if I on purpose create
and use uninitialized variables. Since the gcc version I use is the
same for all architectures, I thought it must have something to do with
compile options (like maybe the option to always initialize stack
variables, or with some gcc plugin), but I have been unable to track it
down.

Guenter