I've begun the task of assessing the 2.6 commits, but the job is unexpectedly large.
What I did:
* Took the output of 'svn log -r60999:66717'. (2.5.2 was branched from rev. 60999, so I'm ignoring commits to the trunk before 2.5.2 was branch, which may miss some things.)
* Wrote a little script to explode the log into a separate file for each commit. This resulted in about 2200 files.
* Did a long series of 'grep' and 'rm' commands to remove irrelevant files. For example, if the commit touches abc.py, bytesobject.c, bytearrayobject.c, etc. it was removed.
* Wrote a little script to find commits that only touch the docs, and moved them aside.
At this point I still have 1191 files left. Many of these commits are still irrelevant, but this is a lot for me to look through. A tarball with the remaining commits is at
Can we parallelize the job? One person could take commits starting with '01', another with '02', etc. Or each person could assess the commits they made.