[lkml]   [2020]   [Oct]   [2]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
Messages in this thread
SubjectRe: [PATCH v4 00/52] Fix html build with Sphinx 3.1 and above
Em Thu, 1 Oct 2020 15:48:54 -0600
Jonathan Corbet <> escreveu:

> On Wed, 30 Sep 2020 15:24:23 +0200
> Mauro Carvalho Chehab <> wrote:
> > This series add proper support for Sphinx 3.1 and above for building the html docs.
> >
> > This series comes after the one I wrote fixing the warnings.
> >
> > With this series applied on the top of next-20200922, there are just 12
> > warnings:
> > - 2 of them happens also on Sphinx 2.4.4
> > (both seem easy to fix: I'll send later fixes for those);
> > - 10 happens only on Sphinx 3.2.1.
> >
> > The new warnings are all due to duplicated C domain cross-reference symbols.
> >
> > Basically, the C domain on Sphinx doesn't allow to have an struct and
> > a function with the same name. I opened an issue on Sphinx.:
> >
> >
> >
> > Hopefully, some newer version may have it fixed.
> >
> > There is still one thing that requires a fix: the The
> > way cross-references are stored with Sphinx 3 are different.
> > I didn't try yet to address it, but I suspect that it shouldn't be
> > hard to address it.
> Modulo my comment on part 4, I think this is what we want. It's kind of
> unfortunate that it's all necessary, but that's the way things are these
> days, I guess.

Yes, it is unfortunate. Yet, my feeling is that this approach provides
us a good way to transition to the new C domain parser, which has some
advantages over the older code.

> This part is a bit intimidating, though:
> > 288 files changed, 1709 insertions(+), 2183 deletions(-)

Yes. Yet, the media userspace API is responsible for most of it:

224 files changed, 1076 insertions(+), 1716 deletions(-)

As most of the uAPI documentation doesn't rely on kernel-doc
(mostly due to historic reasons).

To be fair, most of the other changes outside the building system
are due to already-existing issues at the documentation. There are
several kernel-doc tags creating duplicated C references.

While Sphinx < 3 C domain parser were able to detect them, such
warnings are disabled by default.

> Should we maybe position this as an end-of-merge-window blast, once other
> stuff has hopefully mostly settled? I can certainly warn Linus that it's
> coming when I send the main docs pull.

Yeah, adding this series, together with the remaining patches fixing
all warnings with Sphinx < 3 at the end of the merge window seems
the right thing to do, IMO.

> I wonder how soon we could pull our minimum version forward to 3.1 and
> drop a bunch of stuff? I fear it may not be for a while, alas...

I guess it may take some time. In any case, for the next minimal
version, I would set for the upcoming one that will fix those
issue reported here:

One such example is this warning:

.../Documentation/driver-api/miscellaneous:48: ../drivers/pwm/core.c:669: WARNING: Duplicate C declaration, also defined in 'driver-api/miscellaneous'.
Declaration is 'int pwm_capture (struct pwm_device *pwm, struct pwm_capture *result, unsigned long timeout)'.

(See: we have a function pwm_capture and a struct with the
same name. Both have kernel-doc tags at drivers/pwm/core.c)

Without fixing it, there's no way to produce a zero-warning
docs building[1] with Sphinx versions above 2.4.x.

[1] Well, it could be possible to rename such kAPI symbols to
make Sphinx happy, but IMO, the C domain there should be able
to fully support the C language.

> Thanks for doing all this work,



 \ /
  Last update: 2020-10-02 08:53    [W:0.313 / U:11.172 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site