[Ksummit-discuss] [CORE TOPIC] Reviewing new API/ABI

[Hans Verkuil]

On 05/07/14 15:30, Laurent Pinchart wrote:
> On Wednesday 07 May 2014 14:36:27 Daniel Vetter wrote:
>> On Wed, May 7, 2014 at 12:12 PM, Laurent Pinchart wrote:
>>> We have seen several review, test and documentation procedures being
>>> developed for different subsystems in the recent past (two examples are
>>> the DRM i915 driver API test rules explained by Daniel Vetter in a reply
>>> to this mail thread, and the V4L test suite and documentation procedure)
>>> but I have seen little effort to consolidate good practice rules. The
>>> kernel would certainly benefit from both sharing information about how
>>> various subsystems tackle the API review/test/documentation problem.
>>>
>>> Forcing all subsystems to adhere and enforce a superset of rules would
>>> likely put too much burden on maintainers and developers, especially for
>>> the smaller subsystems. However, I believe we could help by gathering and
>>> consolidating the good practice rules in a single location under
>>> Documentation/. Maintainers could then implement those rules (or a subset
>>> thereof) without having to reinvent the wheel. Rules such as "return
>>> -EINVAL when a reserved parameter is set" are not complex to implement in
>>> code, the real challenge is to implement them in the brain of all
>>> developers and reviewers.
>>
>> I'd be very interested in a discussions about existing best practices
>> already developed in different subsystems and figuring out what
>> minimal standards we should requires across the board.
> 
> So would I. The same topic is being discussed in the "stable issues" mail 
> thread, it thus looks like a good candidate to me. On the V4L side Hans 
> Verkuil (CC'ed) would likely be interested as a core V4L reviewer and 
> userspace test case developer.
> 

Absolutely. I've done a lot of work in that area. The V4L API is very large,
and without compliance tools it is almost impossible to write a faultless
driver. We now have about 95% coverage of the whole API and these days it is
a requirement of new drivers that they pass the test suite, otherwise I
will reject them.

And the flip-side of that is to provide applications with good test drivers
so they can test their application without requiring hard-to-get hardware
if they want to test support for features that are not commonly found.

I developed a very nice test driver for that. It needs a bit more work before
I can upstream it, though.

My experience with creating these tools is that they force you to take a
very critical look at the API. Any ambiguities you missed tend to be exposed
by creating these compliance tests.

Regards,

	Hans