Eric Blake wrote:
> I noticed this during my audit for stderr vs. exec. We've made previous
> cleanups along these lines, but usages have crept back in. OK to commit?
...
> Subject: [PATCH] maint: avoid exiting with magic number
>
> * cfg.mk (sc_prohibit_magic_number_exit)
> (sc_prohibit_error_success): New rules.
> * .x-sc_prohibit_magic_number_exit: Add exemptions.
> * Makefile.am (syntax_check_exceptions): Distribute it.
> * lib/euidaccess-stat.c (main): Fix culprits.
> * src/chcon.c (main): Likewise.
> * src/runcon.c (main): Likewise.
> * src/setuidgid.c (main): Likewise.
Nice!
...
> diff --git a/cfg.mk b/cfg.mk
> index 1e5108b..cce945a 100644
> --- a/cfg.mk
> +++ b/cfg.mk
> @@ -203,6 +203,18 @@ sc_prohibit_stat_macro_address:
> msg='stat() and lstat() may be function-like macros' \
> $(_prohibit_regexp)
>
> +# Pass EXIT_*, not number, to usage, exit, and error
> +sc_prohibit_magic_number_exit:
> + @re='\<(usage|exit) \([0-9]|\<error \([1-9][0-9]*,' \
Please allow 0 or one space between the function name
and the opening parenthesis, so this rule will also be useful
to projects with different code formatting conventions.
> + msg='use EXIT_* values rather than magic number' \
> + $(_prohibit_regexp)
> +
> +# error (0) is preferred over error (EXIT_SUCCESS), since it does not exit
> +sc_prohibit_error_success:
> + @re='\<error \(EXIT_SUCCESS,' \
same here
> + msg='use error(0) rather than error(EXIT_SUCCESS)' \
which means you'll want to obfuscate the above, e.g., to
msg='use error''(0) rather than error(EXIT_SUCCESS)' \
These seem generally useful enough for gnulib's maint.mk.
