Re: kernel-doc -Werror regression

[Mauro Carvalho Chehab <mchehab+huawei@kernel.org>]

Hi Jani,

Em Fri, 09 Jan 2026 13:07:02 +0200
Jani Nikula <jani.nikula@intel.com> escreveu:

> Mauro, Jon -
> 
> It seems to me with the kernel-doc conversion from perl to python we've
> lost a huge amount of -Werror handling. AFAICS none of the warnings from
> kdoc_parser.py using emit_msg() lead to a non-zero exit status from
> kernel-doc.

The culprit was not the conversion, but this change:
 469c1c9eb6c9 ("kernel-doc: Issue warnings that were silently discarded")

which effectively broke most of -W<foo> command line arguments.

I mentioned it back then:
	https://lore.kernel.org/linux-doc/3fnulqi7hfplggfqevab525clikq7fnsnt72lauddzy32sepyq@maux2bgqa3np/

-

After the conversion, the errors detected during the parsing
phase were moved to be displayed/handled at the output phase.
There, when an error occurs, self.errors is incremented at:

	tools/lib/python/kdoc/kdoc_files.py

Then, at kernel-doc.py, we have:

    error_count = kfiles.errors
    if not error_count:
        sys.exit(0)

    if args.werror:
        print("%s warnings as errors" % error_count)    # pylint: disable=C0209
        sys.exit(error_count)

Patch 469c1c9eb6c9 broke it, among other things, as it now emits
errors at parsing time, breaking the logic that inhibits errors
with -W<foo>, and breaking -Werror as well, as the error is not
counted anymore.

That's said, there is another issue there as well, which I hadn't
time yet to fix: we can't do sys.exit(error_count), as the return
code is an 8-bit value.

I'll work on a patch series to address such issues along the week.

Thanks,
Mauro