Re: [PATCH v2] ext4: Enable code path when DX_DEBUG is set

[Nick Desaulniers <ndesaulniers@google.com>]

On Mon, Feb 1, 2021 at 2:48 PM Theodore Ts'o <tytso@mit.edu> wrote:
>
> On Mon, Feb 01, 2021 at 07:05:11PM -0300, Vinicius Tinti wrote:
> >
> > The goal is to try to detect real bugs. In this instance specifically I
> > suggested to remove the "if (0) {...}" because it sounded like an
> > unused code.
> >
> > If it is useful it is fine to keep.
>
> The trick was that it was unused code, but it was pretty obviously
> deliberate, which should have implied that at some point, it was
> considered useful.   :-)
>
> It was the fact that you were so determined to find a way to suppress
> the warning, suggesting multiple tactics, which made me wonder.... why
> were you going through so much effort to silence the warning if the
> goal was *not* to turn it on unconditionally everywhere?

Because a maintainer might say "oh, I meant to turn that back on years
ago" or "that should not have been committed!"  Hasn't happened yet,
doesn't mean it's impossible.  Vinicius asked how he can help. I said
"go see if any instances of this warning are that case."

>
> I suspect the much more useful thing to consider is how can we suggest
> hueristics to the Clang folks to make the warning more helpful.  For
> example, Coverity will warn about the following:
>
> void test_func(unsigned int arg)
> {
>         if (arg < 0) {
>                 printf("Hello, world\n")
>         }
> }

Put that code in in godbolt.org (https://godbolt.org/z/E7KK9T) and
you'll see that both compilers already warn here on -Wextra (via
-Wtautological-unsigned-zero-compare in clang or -Wtype-limits in
GCC).
clang:

warning: result of comparison of unsigned expression < 0 is always
false [-Wtautological-unsigned-zero-compare]
        if (arg < 0) {
            ~~~ ^ ~

gcc:

warning: comparison of unsigned expression in '< 0' is always false
[-Wtype-limits]
    3 |         if (arg < 0) {
      |                 ^


>
> P.S.  If anyone wants to file a feature request bug with the Clang
> developers, feel free.  :-)



-- 
Thanks,
~Nick Desaulniers