Some 5.14 development statistics

Posted Aug 31, 2021 20:02 UTC (Tue) by ndesaulniers (subscriber, #110768)
Parent article: Some 5.14 development statistics

I miss the stats on reviewed-by and tested-by tags. Those are meaningful ways of contributing to the kernel that aren't necessarily writing code. And given the rate of change of the codebase, these contributions are meaningful and welcome.

Review and test tags

Posted Aug 31, 2021 20:23 UTC (Tue) by corbet (editor, #1) [Link] (8 responses)

I can certainly put them back; I'll try to post them as a comment here, but that will require waiting until a blessed future moment when my office isn't overrun by window-replacement contractors.

In general I've kind of lost my enthusiasm for those stats, though. I'm not convinced that they give anything close to an accurate picture of the work that is going on in the community.

Review and test tags

Posted Aug 31, 2021 21:03 UTC (Tue) by willy (subscriber, #9762) [Link]

I don't think they're even close to accurate or meaningful, but there's a good endorphin hit for the people who contribute them, and that encourages more of them, which we do want.

Review and test tags

Posted Sep 1, 2021 6:26 UTC (Wed) by pbonzini (subscriber, #60935) [Link]

What about R-b and T-b stats by subsystem? I am especially curious of how many reviews come from designated reviewers and how many are from "bystanders".

Review and test tags

Posted Sep 1, 2021 18:13 UTC (Wed) by edeloget (subscriber, #88392) [Link]

> ... when my office isn't overrun by window-replacement contractors.

They're installing veluxes ?

Review and test tags

Posted Sep 1, 2021 18:57 UTC (Wed) by corbet (editor, #1) [Link] (3 responses)

As promised, here are the missing stats.

Developers with the most reviews (total 6408)
Rob Herring	172	2.7%
Hannes Reinecke	121	1.9%
Andy Shevchenko	110	1.7%
Felix Kuehling	104	1.6%
Christian König	99	1.5%
Christoph Hellwig	94	1.5%
Andrew Lunn	92	1.4%
Darrick J. Wong	89	1.4%
Hans de Goede	79	1.2%
Linus Walleij	79	1.2%
Alex Deucher	77	1.2%
Guenter Roeck	74	1.2%
Bjorn Andersson	68	1.1%
Himanshu Madhani	67	1.0%
Huang Rui	61	1.0%
Hawking Zhang	60	0.9%
Dan Carpenter	59	0.9%
Lijo Lazar	59	0.9%
Oded Gabbay	58	0.9%
Chaitanya Kulkarni	55	0.9%

Developers with the most test credits (total 956)
Daniel Wheeler	154	16.1%
Frieder Schrempf	41	4.3%
Tony Brelinski	39	4.1%
Vincent Guittot	23	2.4%
Dvora Fuxbrumer	22	2.3%
Anand Jain	21	2.2%
Hongyu Ning	20	2.1%
Ritesh Harjani	19	2.0%
Don Hiatt	19	2.0%
Marc Zyngier	14	1.5%
Benjamin Li	12	1.3%
Jernej Skrabec	10	1.0%
Dave Switzer	10	1.0%
Daniel Almeida	10	1.0%
James Clark	9	0.9%
Kris Pan	9	0.9%
Guenter Roeck	8	0.8%
Arnaldo Carvalho de Melo	8	0.8%
Sachin Sant	7	0.7%

Developers with the most report credits (total 1216)
kernel test robot	233	19.2%
Hulk Robot	212	17.4%
Abaci	88	7.2%
Syzbot	60	4.9%
Dan Carpenter	56	4.6%
Stephen Rothwell	22	1.8%
Pierre-Louis Bossart	12	1.0%
Will Deacon	11	0.9%
Guenter Roeck	10	0.8%
Nathan Chancellor	10	0.8%
Sachin Sant	8	0.7%
Colin Ian King	7	0.6%
Alexey Kardashevskiy	7	0.6%
Marek Szyprowski	6	0.5%
coverity-bot	6	0.5%
Mauro Carvalho Chehab	5	0.4%
Geert Uytterhoeven	4	0.3%
Tom Rix	4	0.3%
Randy Dunlap	4	0.3%
Matthew Wilcox	4	0.3%
Johannes Berg	4	0.3%
Naresh Kamboju	4	0.3%

Review and test tags

Posted Sep 2, 2021 2:55 UTC (Thu) by pabs (subscriber, #43278) [Link] (2 responses)

Hmm, there are a lot less people doing testing/reporting than doing reviews.

Review and test tags

Posted Sep 2, 2021 14:22 UTC (Thu) by corbet (editor, #1) [Link] (1 responses)

Either that, or there are a lot less Tested-by and Reported-by credits being given; as noted before, I don't think these numbers reflect the actual level of activity very well.

Review and test tags

Posted Sep 6, 2021 8:37 UTC (Mon) by geert (subscriber, #98403) [Link]

Indeed. Examples of typical activities leading to missing credits:
1. Reviewed-by:
- Pointing out issues, leading to v${N+1},
- Taking a patch through your tree, adding a Signed-of-by: instead of a Reviewed-by.
2. Tested-by:
- Pointing out issues, leading to v${N+1},
3. Reported-by:
- Fixing an issue you detected yourself.
Only the last one can be extracted from git, by looking at Fixes: tags.
The rest could be extracted from lore though, but would involve much more work.

Review and test tags

Posted Sep 6, 2021 17:34 UTC (Mon) by kdave (subscriber, #44472) [Link]

I'm not sure that accuracy is the point of all those stats anyway, for example it's easy to be among top 10 with Signed-off just by sending typo fixes or trivial code changes. That's going for quantity, while reviewed-by is tracking how many people care about quality. Stating the obvious, reviews are always on somebody else's code. That's similar to the non-author signed-off by, that's the maintainer's default job, but there's no mandate to do reviews. It's voluntary and partly taking responsibility for the quality.

"Not enough reviews" used to be a common topic on conferences, nobody has a universally working recipe how to get more reviewers. Pushing people to do that does not work, that would hardly improve quality.

Keeping the stats at least on the release overview page makes it a bit more visible as a thing that the community cares about. You can imagine some poor soul not sending a single patch but reviewing everything, thus making sure the end result is ok, yet this is somehow "forgotten" to be mentioned.