Who reverted 2.6.25 - stats

Who reverted 2.6.25 - stats

[Jan Engelhardt]

(Subject modeled upon the "Who wrote 2.6.x" series from LWN)
Hello everyone,


following the implied "suggestion" of David Miller to track reverts in 
http://marc.info/?l=linux-kernel&m=120959059828048&w=2 , I stitched 
together a short script that evaluates the commit log summary and prints 
a overview who has the most reverts and who got reverted most. There are 
surprisingly few revert commits -- currently, perhaps because, as he 
says, a "reluctance to even suggest reverts".

One thing I noticed is that reverts are not always marked with the word 
"revert" in the summary line even though it normally would be, for 
example
	6b8e1c7e x86: 8K stacks by default
	d61ecf0b x86: 4kstacks default
(People evading the revert statistics?)
So that is not counted, but I wanted to make aware of it.
Stats are below.

> git log v2.6.25..linus/master --pretty="format:<%an>%s" | ./git-revert-stats
 P#  COMM#  NAME --- Top people doing the revert:
  1      1  Paul Mackerras
  2      1  David S. Miller
  3      1  Ingo Molnar
  4      1  Andrew Morton
  5      1  John W. Linville

 P#  COMM#  NAME --- Top developers with reverted commits:
  1      1  Andreas Schwab
  2      1  Michael Beasley
  3      1  Ingo Molnar
  4      1  Luis Carlos Cobo
  5      1  Yasunori Goto

A bit more interesting are the tallies for older releases:

> git log v2.6.24..v2.6.25 --pretty="format:<%an>%s" | ./git-revert-stats -o
 P#  COMM#  NAME --- Top people doing the revert:
  1      6  Linus Torvalds
  2      5  Len Brown
  3      4  Andrew Morton
  4      3  Ingo Molnar
  5      2  Paul Mackerras
  6      2  David S. Miller
  7      2  Bartlomiej Zolnierkiewicz
  *     10 developers with only one revert are omitted

 P#  COMM#  NAME --- Top developers with reverted commits:
  1      2  Matthew Garrett
  2      2  Ingo Molnar
  3      2  Alexey Starikovskiy
  4      2  Christoph Lameter
  *     20 developers with only one revert are omitted

> git log v2.6.23..v2.6.24 --pretty="format:<%an>%s" | ./git-revert-stats -o
 P#  COMM#  NAME --- Top people doing the revert:
  1     13  Linus Torvalds
  2      7  Andrew Morton
  3      5  Len Brown
  4      2  Jens Axboe
  5      2  Thomas Gleixner
  6      2  David Woodhouse
  *     11 developers with only one revert are omitted

 P#  COMM#  NAME --- Top developers with reverted commits:
  1      2  Ingo Molnar
  2      2  Alexey Starikovskiy
  3      2  Stephen Hemminger
  4      2  Mathieu Desnoyers
  *     25 developers with only one revert are omitted

> git log v2.6.22..v2.6.23 --pretty="format:<%an>%s" | ./git-revert-stats -o
 P#  COMM#  NAME --- Top people doing the revert:
  1     13  Linus Torvalds
  2      5  Andrew Morton
  *     11 developers with only one revert are omitted

 P#  COMM#  NAME --- Top developers with reverted commits:
  1      2  Thomas Gleixner
  2      2  Ryusuke Konishi
  *     19 developers with only one revert are omitted

> git log v2.6.21..v2.6.22 --pretty="format:<%an>%s" | ./git-revert-stats -o
 P#  COMM#  NAME --- Top people doing the revert:
  1      7  Linus Torvalds
  2      3  Paul Mackerras
  3      3  Len Brown
  4      2  Andi Kleen
  5      2  Andrew Morton
  *      5 developers with only one revert are omitted

 P#  COMM#  NAME --- Top developers with reverted commits:
  1      2  Stephen Rothwell
  2      2  Jake Moilanen
  *     10 developers with only one revert are omitted

Find the script at 
http://dev.medozas.de/gitweb.cgi?p=hxtools;a=blob;f=libexec/git-revert-stats;hb=master
(it is not yet in a hxtools tarball).

Re: Who reverted 2.6.25 - stats

[Stefan Richter]

Jan Engelhardt wrote:
> One thing I noticed is that reverts are not always marked with the word 
> "revert" in the summary line even though it normally would be,

IMO the summary line of a fix should tell what is being fixed.  The
changelog can then go on to explain how the fix works.  For example, by
reverting all or parts of a previous commit.
-- 
Stefan Richter
-=====-==--- -=-= ---=-
http://arcgraph.de/sr/

Re: Who reverted 2.6.25 - stats

[Diego Calleja]

El Fri, 2 May 2008 14:16:19 +0200 (CEST), Jan Engelhardt <jengelh@medozas.de> escribió:

> following the implied "suggestion" of David Miller to track reverts in 
> http://marc.info/?l=linux-kernel&m=120959059828048&w=2 , I stitched 
> together a short script that evaluates the commit log summary and prints 
> a overview who has the most reverts and who got reverted most. There are 
> surprisingly few revert commits -- currently, perhaps because, as he 
> says, a "reluctance to even suggest reverts".

But what exactly those numbers mean?

A big number of reverts can mean that a developer usually submits unstable
stuff. But it can also mean that a developer cares so much about stability
that he'd rather revert, fix it and try to merge it in the next development
cycle. The "bad" maintainer that submits unstable stuff may try to fix
the reported bug instead of reverting it (but he won't fix the other 
undiscovered bugs that are still there because the code was not tested
well enought)

I don't think those numbers are very useful. Even the "number of bug reports
filed against the code of a given developer" is not considered meaningful as
measurement of good programming.