LWN: Comments on "Some 4.12 development statistics" https://lwn.net/Articles/726950/ This is a special feed containing comments posted to the individual LWN article titled "Some 4.12 development statistics". en-us Fri, 19 Sep 2025 16:26:18 +0000 Fri, 19 Sep 2025 16:26:18 +0000 https://www.rssboard.org/rss-specification lwn@lwn.net Metrics https://lwn.net/Articles/729462/ https://lwn.net/Articles/729462/ andy_shev <div class="FormattedComment"> Probably I didn't put it clear, I meant independent "modules / drivers" under "files", and statistics here is "by changesets". <br> <p> $ gt4tree.sh v4.12..origin/master<br> Christoph Hellwig (246):<br> Mauro Carvalho Chehab (169):<br> Thomas Gleixner (143):<br> Takashi Iwai (130):<br> Chris Wilson (126):<br> Arvind Yadav (122):<br> ...<br> <p> See the last name in the list, here. While I appreciate the work done by Arvind, it's still a mechanical routine which can be easily scripted (coccinelle, sed, etc).<br> </div> Tue, 01 Aug 2017 16:21:10 +0000 Metrics https://lwn.net/Articles/729288/ https://lwn.net/Articles/729288/ corbet As far as I've been able to tell, the results that come out of cregit closely match the (simpler) "by lines" metric. <p> Mon, 31 Jul 2017 13:42:33 +0000 Metrics https://lwn.net/Articles/729277/ https://lwn.net/Articles/729277/ pabs <div class="FormattedComment"> Maybe something based on tokens using cregit?<br> </div> Mon, 31 Jul 2017 02:55:19 +0000 Metrics https://lwn.net/Articles/729262/ https://lwn.net/Articles/729262/ corbet Nothing I've done here counts the number of affected files, so not necessarily. But I'll freely admit that the metrics used here — changesets and changed lines — are awful. I'd sure like to find something better if anybody has any good ideas. Sun, 30 Jul 2017 19:09:58 +0000 Some 4.12 development statistics https://lwn.net/Articles/729251/ https://lwn.net/Articles/729251/ andy_shev <div class="FormattedComment"> I didn't check myself so the following question. Imagine that a maintainer does <br> <p> % sed -i -e 's/old_api_call/new_api_call/g' &lt;list of affected files&gt;<br> <p> If number of list of affected files greate than let's say 100 she or he immediatelly poped up in the statistics, right? But it's not what we would like to see there for my point of view.<br> <p> </div> Sun, 30 Jul 2017 15:31:32 +0000 Some 4.12 development statistics https://lwn.net/Articles/727382/ https://lwn.net/Articles/727382/ corbet Yes, that's a manual fix for $REASONS; it got done then undone when I fixed an unrelated problem. Fixed again. Fri, 07 Jul 2017 16:25:32 +0000 Some 4.12 development statistics https://lwn.net/Articles/727379/ https://lwn.net/Articles/727379/ kdave <div class="FormattedComment"> 'Novell' as an employer does not exist anymore, s/Novell/SUSE/.<br> </div> Fri, 07 Jul 2017 15:12:56 +0000 Some 4.12 development statistics https://lwn.net/Articles/727329/ https://lwn.net/Articles/727329/ broonie <div class="FormattedComment"> It's common to get very big line counts by doing things like removing old drivers or adding new devices with large register sets where you can often get big machine generated sets of register definitions and/or defaults.<br> </div> Thu, 06 Jul 2017 19:19:30 +0000 Some 4.12 development statistics https://lwn.net/Articles/727208/ https://lwn.net/Articles/727208/ dankohn <div class="FormattedComment"> For posterity, Jon Corbet responded to me by email to explain:<br> <p> In short, to get the numbers for (say) Intel, I do a normal-looking gitdm<br> run but add:<br> <p> -C "Intel" -f intel.files<br> <p> That will (1) only look at commits from an Intel employee, and (2) put a<br> list of touched files in "intel.files". A "sort -rn" on that file will<br> put them into decreasing order, at which point the numbers are relatively<br> easy to pick out.<br> </div> Wed, 05 Jul 2017 18:10:46 +0000 Some 4.12 development statistics https://lwn.net/Articles/727147/ https://lwn.net/Articles/727147/ arnd <div class="FormattedComment"> I made a chart showing the git trees that contributed most to the kernel over time, by number of patches and number of lines:<br> <p> <a href="https://docs.google.com/spreadsheets/d/15lFZZ_RygzGgwJSPc6PITm9kfZlFIDwFeoaR1GvLjyo/edit?usp=sharing">https://docs.google.com/spreadsheets/d/15lFZZ_RygzGgwJSPc...</a><br> <p> In 4.12 both staging and drm were really big compared to the rest. We have had previous releases in which one of them dominated the total patch count or diff size but not both at the same time, and we have other trees that are relatively big (net-next, tip, arm-soc, ...) but more constant.<br> </div> Wed, 05 Jul 2017 09:32:53 +0000 Some 4.12 development statistics https://lwn.net/Articles/727145/ https://lwn.net/Articles/727145/ gby <div class="FormattedComment"> They most probably didn't *really* read many of those lines and almost for sure did not author many of them. For example, most of the 17,593 lines listed under my name were written by others. I'm just the one upstreaming the driver and maintaining it. Tons of that code tends to be a long include file full of register and descriptor definitions, quite often auto-generated from hardware definition files. <br> </div> Wed, 05 Jul 2017 07:11:59 +0000 Some 4.12 development statistics https://lwn.net/Articles/727142/ https://lwn.net/Articles/727142/ dgm <div class="FormattedComment"> I find it astonishing that Alex Deucher and Alan Cox have handled to touch more than half a million lines of code between them two. I'm not sure if I would be able to *read* that many lines, much less change them.<br> </div> Wed, 05 Jul 2017 06:24:17 +0000 Some 4.12 development statistics https://lwn.net/Articles/727140/ https://lwn.net/Articles/727140/ pabs <div class="FormattedComment"> This is definitely the first time I've seen it on LWN.<br> </div> Wed, 05 Jul 2017 05:48:27 +0000 Some 4.12 development statistics https://lwn.net/Articles/727128/ https://lwn.net/Articles/727128/ dankohn <div class="FormattedComment"> Is the directory per company info a new feature? Could you please say how to generate it from gitdm?<br> </div> Tue, 04 Jul 2017 19:52:35 +0000