Re: Lint-related patches for coreutils

[Pádraig Brady]

On 31/01/2022 20:16, Paul Eggert wrote:
> I went through a lint pass for Coreutils and installed the attached
> patches. These fix only one or two real (albeit unlikely and/or
> technical) bugs. Most of them remove "#ifdef lint"s or "IF_LINT"s that
> should no longer be needed with modern GCC and runtimes, the idea being
> to lessen the difference between linted and unlinted code.

Note if there is no functional change, it's better to use
a git summary line of the form "maint: ..."
(and also squash similar patches if possible)

> If you're
> building with older GCC and/or non-GCC you may run into a few glitches,
> which you can ignore with "make WERROR_CFLAGS=" or with "./configure
> --disable-gcc-warnings" (the default unless you're building from Git).
>
> The biggest lint-oriented change here is the new macro 'main_exit',
> which should be used only in main functions and is equivalent to either
> 'exit' or 'return', depending on whether we're pacifying gcc
> -fsanitize=leak. Using this macro helped me to eliminate a lot of
> 'IF_LINT (free (x))' calls.

We used returns to work with single binary mode.

$ ./configure --enable-single-binary --quiet
$ make

src/expr.c: In function 'single_binary_main_expr':
error: function might be candidate for attribute 'noreturn'
  [-Werror=suggest-attribute=noreturn]
note: in definition of macro 'main'