Re: [PATCH 04/14] ABI: better identificate tables

From: Johan Hovold <johan@kernel.org>
To: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Mauro Carvalho Chehab <mchehab+samsung@kernel.org>,
	Johan Hovold <johan@kernel.org>,
	Linux Doc Mailing List <linux-doc@vger.kernel.org>,
	Mauro Carvalho Chehab <mchehab@s-opensource.com>,
	Mauro Carvalho Chehab <mchehab@infradead.org>,
	linux-kernel@vger.kernel.org, Jonathan Corbet <corbet@lwn.net>,
	Stefan Achatz <erazor_de@users.sourceforge.net>
Subject: Re: [PATCH 04/14] ABI: better identificate tables
Date: Thu, 20 Jun 2019 14:01:50 +0200	[thread overview]
Message-ID: <20190620120150.GH6241@localhost> (raw)
In-Reply-To: <20190619150207.GA19346@kroah.com>

On Wed, Jun 19, 2019 at 05:02:07PM +0200, Greg Kroah-Hartman wrote:
> On Wed, Jun 19, 2019 at 10:56:33AM -0300, Mauro Carvalho Chehab wrote:
> > Hi Johan,
> > 
> > Em Wed, 19 Jun 2019 14:51:35 +0200
> > Johan Hovold <johan@kernel.org> escreveu:
> > 
> > > On Thu, Jun 13, 2019 at 11:04:10PM -0300, Mauro Carvalho Chehab wrote:
> > > > From: Mauro Carvalho Chehab <mchehab@s-opensource.com>
> > > > 
> > > > When parsing via script, it is important to know if the script
> > > > should consider a description as a literal block that should
> > > > be displayed as-is, or if the description can be considered
> > > > as a normal text.
> > > > 
> > > > Change descriptions to ensure that the preceding line of a table
> > > > ends with a colon. That makes easy to identify the need of a
> > > > literal block.  
> > > 
> > > In the cover letter you say that the first four patches of this series,
> > > including this one, "fix some ABI descriptions that are violating the
> > > syntax described at Documentation/ABI/README". This seems a bit harsh,
> > > given that it's you that is now *introducing* a new syntax requirement
> > > to assist your script.
> > 
> > Yeah, what's there at the cover letter doesn't apply to this specific
> > patch. The thing is that I wrote this series a lot of time ago (2016/17).

Got it, thanks.

[...]

> > In the specific case of this patch, the ":" there actually makes sense
> > for someone that it is reading it as a text file, and it is an easy
> > hack to make it parse better.

Human readers probably depend on more on tabulation and white space.
When the preceding description wasn't using a colon to begin with (and
you just replace s/\./:/) it can even look weird, but no big deal.

> > > Specifically, this new requirement isn't documented anywhere AFAICT, so
> > > how will anyone adding new ABI descriptions learn about it?
> > 
> > Yeah, either that or provide an alternative to "Description" tag, to be
> > used with more complex ABI descriptions.
> > 
> > One of the things that occurred to me, back on 2017, is that we should
> > have a way to to specify that an specific ABI description would have
> > a rich format. Something like:

[...]

> I don't know when "Description" and "RST-Description" would be used.
> Why not just parse "Description" like rst text and if things are "messy"
> we fix them up as found, like you did with the ":" here?  It doesn't
> have to be complex, we can always fix them up after-the-fact if new
> stuff gets added that doesn't quite parse properly.
> 
> Just like we do for most kernel-doc formatting :)

But kernel-doc has a documented format, which was sort of the point I
was trying to make. If the new get_abi.pl scripts expects a colon I
think it should be mentioned somewhere (e.g. Documentation/ABI/README).

Grepping for attribute entries in linux-next still reveals a number
descriptions that still lack that colon and use varying formatting. More
are bound to be added later, but perhaps that's ok depending on what
you're aiming at here.

Johan