bpf.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Grant Seltzer Richman <grantseltzer@gmail.com>
To: Andrii Nakryiko <andrii.nakryiko@gmail.com>
Cc: bpf <bpf@vger.kernel.org>
Subject: Re: Generating libbpf API documentation
Date: Wed, 21 Apr 2021 15:24:13 -0400	[thread overview]
Message-ID: <CAO658oXWP2dsAsD9S8=EjosnoE4ND6LWr3js_yiDNGNw3ZqGQg@mail.gmail.com> (raw)
In-Reply-To: <CAEf4BzbxZsS+6S+qdXHFSYPGsevQhxdXqsVH8Z9HGTWreP5uGQ@mail.gmail.com>

On Tue, Apr 20, 2021 at 12:26 AM Andrii Nakryiko
<andrii.nakryiko@gmail.com> wrote:
>
> On Fri, Apr 16, 2021 at 12:38 PM Grant Seltzer Richman
> <grantseltzer@gmail.com> wrote:
> >
> > On Tue, Mar 16, 2021 at 4:14 PM Grant Seltzer Richman
> > <grantseltzer@gmail.com> wrote:
> > >
> > > On Mon, Mar 15, 2021 at 8:47 PM Andrii Nakryiko
> > > <andrii.nakryiko@gmail.com> wrote:
> > > >
> > > > On Mon, Mar 15, 2021 at 9:51 AM Grant Seltzer Richman
> > > > <grantseltzer@gmail.com> wrote:
> > > > >
> > > > > Hi all,
> > > > >
> > > > > I have been experimenting with ways to contribute documentation to
> > > > > libbpf to make it easier for developers of bpf projects to use it.
> > > > > With the goal of making a documentation site that is easy to
> > > > > maintain/generate I found Doxygen (many of you may have experience
> > > > > with it, I did not). I set up a CI/CD workflow using github actions
> > > > > that runs doxygen on the libbpf mirror hosted there, and hosts the
> > > > > produced HTML using netlify. You can find the currently hosted version
> > > > > of it at https://libbpf-docs.netlify.app (I would gladly donate a real
> > > > > domain name for this purpose). The docs generation workflow is in my
> > > > > github repo here: https://github.com/grantseltzer/libbpf-docs
> > > >
> > > > Thanks for investigating this! I've look at libbpf-docs.netlify.app,
> > > > and it seems like it just contains a list of structs and their fields
> > > > (both those that are part of libbpf API, as well as internal). Out of
> > > > all functions only two are listed there (libbpf_nla_parse_nested and
> > > > libbpf_nla_parse) and both are not part of libbpf API as well. So I
> > > > understand that I don't see any comments due to the '/**' format
> > > > (though it would be easy to run sed script adding it everywhere, just
> > > > as part of an experiment), but I'm not sure why none of API functions
> > > > are present there?
> > > >
> > > > I think kernel docs used to be hosted on readthedocs.org, seems like
> > > > they are also providing hosting for open-source projects, so that
> > > > would solve the problem of the hosting. Have you looked at that
> > > > solution? It definitely has a bit more modern UI that
> > > > Doxygen-generated one :) but I don't know what are the real
> > > > differences between Sphinx and Doxygen and which one we should choose.
> > > >
> > > > >
> > > > > In order to make this work all we would need is to format comments
> > > > > above functions we want to document. Doxygen requires that the comment
> > > > > just be in a block that starts with `/**`. I don't think doxygen
> > > > > specific directives should be committed to code but I think this is a
> > > > > fine convention to follow. Other doxygen directives (i.e. having
> > > > > `@file` in every file) can be faked using a step I have in the github
> > > > > actions.
> > > > >
> > > > > What does everyone think? Can we agree on this convention and start
> > > > > contributing documentation in this way? Any pitfalls to doxygen I'm
> > > > > not familiar with?
> > > > >
> > > > > Thanks!
> > >
> > > As far as I understand Doxygen's only criteria for generating
> > > documentation for functions is if the correctly formatted comment is
> > > present. I've changed the repo that the libbpf-docs.netlify.app
> > > website uses to track a fork libbpf I have on my personal account. I
> > > added comments above some ringbuffer functions to demonstrate this.
> > >
> > > Interestingly the two functions that already show up
> > > (libbpf_nla_parse/parse_nested) have comments which are specifically
> > > formatted for doxygen, even including directives for arguments and
> > > related functions.
> > >
> > > I have heard of Sphinx/read-the-docs but didn't look too deeply into
> > > it, I'll check it out and report back with my findings!
> >
> > I've finally gotten a chance to circle around to this. I investigated
> > Sphinx and read the docs. As far as I can tell Doxygen is still
> > required for generating that docs from code. Sphinx seems to typically
> > be used to transform markdown documentation files into themed html
> > pages. Sphinx would also enable us to host the documentation on
> > readthedocs's, but it would still be the output of Doxygen, meaning it
> > wouldn't have the nice theme that you see on other readthedocs pages.
> >
> > I have a barebones example set up of what that would look like at
> > libbpf.readthedocs.io which pulls from my fork of the github mirror
> > here: github.com/grantseltzer/libbpf
> >
> > The advantage of this approach is only having free hosting and having
> > a 'readthedocs.io' domain. It would still require CI for pulling in
> > libbpf releases, appending doxygen directives, and of course
> > committing comments in code next to api functions/types.
> >
>
> I didn't have much time to investigate Sphinx vs Doxygen. Reding [0]
> diagonally, seems like you need few extensions (breathe and
> sphinx_rtd_theme) to make everything work. It also seems like
> readthedocs will be able to automatically pull and generate
> documentation, so if all that is true, it still seems like Sphinx +
> readthedocs is the better and more modern approach.
>
>   [0] https://devblogs.microsoft.com/cppblog/clear-functional-c-documentation-with-sphinx-breathe-doxygen-cmake/

That link proved helpful. I was not using the breathe plugin
directives correctly in the previous iteration. Thanks!

>
> > I prefer the previous approach (github actions + netlify/github pages)
> > but regardless would happily set this up if we can start an initiative
> > to add those code comments in code, which I'd also like to help
> > contribute to. I'd also be happy to hear of suggestions of free/open
> > source alternatives for CI.
>
> We currently use Travis CI for libbpf CI, but I'm not very happy with
> it and ideally we should move to GitHub Actions or something along
> those lines.

I recently set up some github actions workflow for the project I help
maintain and really like it so far, I would be happy to help
transition.

>
> >
> > Andrii, do you run the libbpf github org and mirror repo?
>
> Yes, I have admin access along Alexei and Daniel. So we'll be able to
> set up whatever needs to be set up.

I just pushed changes for libbpf.readthedocs.io for you to check out.
The 'API' page has the auto generated docs based on public
functions/structs/enums in libbpf.h. There's a couple of functions
that I added bogus test comments to show what documentation  would
look like. (`libbpf_num_possible_cpus` has a good example). Also the
'BPF Program Types' page is just to serve as an example of how we can
include documentation that isn't just auto generated from code.

I need to read the `sync-kernel.sh` script to better understand how
the mirror works, but after that would it be helpful to open a github
PR? Once that'd get merged I'd transfer the readthedocs libbpf page to
track it. I also want to discuss this on the linux-doc mailing list
for input.

  reply	other threads:[~2021-04-21 19:24 UTC|newest]

Thread overview: 8+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2021-03-15 16:50 Generating libbpf API documentation Grant Seltzer Richman
2021-03-16  0:47 ` Andrii Nakryiko
2021-03-16 20:14   ` Grant Seltzer Richman
2021-04-16 19:38     ` Grant Seltzer Richman
2021-04-20  4:26       ` Andrii Nakryiko
2021-04-21 19:24         ` Grant Seltzer Richman [this message]
2021-04-22  3:58           ` Andrii Nakryiko
2021-04-22 21:20             ` Grant Seltzer Richman

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to='CAO658oXWP2dsAsD9S8=EjosnoE4ND6LWr3js_yiDNGNw3ZqGQg@mail.gmail.com' \
    --to=grantseltzer@gmail.com \
    --cc=andrii.nakryiko@gmail.com \
    --cc=bpf@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).