[yocto-ab] YP Advisory Board: raw metrics

[Philip Balister]

On 10/24/2017 12:04 PM, Nicolas Dechesne wrote:
> On Tue, Oct 24, 2017 at 12:00 PM, Philip Balister <philip at balister.org>
> wrote:
> 
>> On 10/24/2017 09:36 AM, Nicolas Dechesne wrote:
>>> hi Jeff,
>>>
>>> thanks for showing this and sharing the link.
>>>
>>> If we take a look at the top repo, we see
>>>
>>> [image: Inline image 1]
>>>
>>> Commits are counted once (of course), and since several of these trees
>> are
>>> 'combined' together, most commits end up (wrongly?) in poky-contrib. We
>>> probably want to show that in a different way. I can guess that the trees
>>> are processed in a specific 'order' and that each commit is attributed
>> to a
>>> repo when it is first met. So that means that the ordered list we provide
>>> is important, and we need to decide how we want to look at the data.
>>>
>>> There are 2 possible orders (for the main repo):
>>> 1. poky
>>> 2. bitbake
>>> 3. oe-core
>>> 4. meta-yocto
>>> 5. *-contrib
>>>
>>> or
>>>
>>> 1. oe-core
>>> 2. bitbake
>>> 3. meta-yocto
>>> 4. poky
>>> 5. *-contrib
>>>
>>> Both options should give different views. Based on the structure of our
>>> development process, I suppose that the 2nd one makes more sense.
>>>
>>> Some additional questions:
>>>
>>> * Can we choose which branches in repo are processed? At the very least
>>> -next branches should not be taken into account.
>>> * Should we try to track only 'release' branches as much as possible
>>> * Isn't poky-buildhistory irrelevant here? I think we should remove it
>>
>> poky is built from several layers that are already include, so no need
>> to include it. I agree with Nicolas.
>>
>>> * Should we keep the -contrib trees?
>>> * can you share the whole list of repo that are being used?
>>>
>>
>> Layers should now include information about what layers they depend on.
>> By counting the number of times a layer is used by other layers gives us
>> a "popularity" figure of merit. We should look at metrics for popular
>> layers. (I suspect we can figure this out by inspection also)
>>
>> We should also look at including member bsp layers hosts on
>> git.yoctoproject.org.
>>
> 
> they are already. I only copy/paste the top repos, from the website you can
> download the 'top 50' repos, see full list in the attached file. The Web UI
> says there are 86 repos tracked.

Ah thanks for finding that. It seems like someone that understands the
repos these should prune a few more.

Philip

> 
> 
>>
>> Philip
>>
>>
>>
>>> cheers
>>> nico
>>>
>>>
>>> On Tue, Oct 24, 2017 at 7:55 AM, Jeff Osier-Mixon <jefro at jefro.net>
>> wrote:
>>>
>>>> These are the metrics links we discussed yesterday. Think of these as
>>>> building blocks to provide data. Now we need to figure out good
>> questions
>>>> to ask to determine how to measure the health of the community. This
>> will
>>>> be an ongoing effort over the next quarter.
>>>>
>>>> Bitergia is a company that LF has contracted with to provide metrics for
>>>> all of their collaborative projects. Intel paid for additional data sets
>>>> and support on behalf of the Yocto Project. Through these tools we have
>>>> visualized data for the git repos, mailing lists, and bugzilla instance.
>>>> These visualizations are highly configurable, and the whole setup is
>> built
>>>> with open source software.
>>>>
>>>> https://yoctoproject.biterg.io
>>>>
>>>> OpenHub is a BlackDuck project that provides a medium-depth dive into
>> the
>>>> git server.
>>>>
>>>> https://www.openhub.net/p/YoctoProject
>>>>
>>>> Now that we have these tools, I look forward to working with those
>>>> interested on customizing the data set definitions so we know exactly
>> what
>>>> we are looking at.
>>>>
>>>> --
>>>> Jeff Osier-Mixon - Open Source Community Manager, Intel Corporation
>>>>
>>>> --
>>>> _______________________________________________
>>>> yocto-ab mailing list
>>>> yocto-ab at yoctoproject.org
>>>> https://lists.yoctoproject.org/listinfo/yocto-ab
>>>>
>>>>
>>>
>>>
>>>
>>
>