1.. SPDX-License-Identifier: GPL-2.0
2
3=====================================
4Handling messy pull-request diffstats
5=====================================
6
7Subsystem maintainers routinely use ``git request-pull`` as part of the
8process of sending work upstream.  Normally, the result includes a nice
9diffstat that shows which files will be touched and how much of each will
10be changed.  Occasionally, though, a repository with a relatively
11complicated development history will yield a massive diffstat containing a
12great deal of unrelated work.  The result looks ugly and obscures what the
13pull request is actually doing.  This document describes what is happening
14and how to fix things up; it is derived from The Wisdom of Linus Torvalds,
15found in Linus1_ and Linus2_.
16
17.. _Linus1: https://lore.kernel.org/lkml/CAHk-=wg3wXH2JNxkQi+eLZkpuxqV+wPiHhw_Jf7ViH33Sw7PHA@mail.gmail.com/
18.. _Linus2: https://lore.kernel.org/lkml/CAHk-=wgXbSa8yq8Dht8at+gxb_idnJ7X5qWZQWRBN4_CUPr=eQ@mail.gmail.com/
19
20A Git development history proceeds as a series of commits.  In a simplified
21manner, mainline kernel development looks like this::
22
23  ... vM --- vN-rc1 --- vN-rc2 --- vN-rc3 --- ... --- vN-rc7 --- vN
24
25If one wants to see what has changed between two points, a command like
26this will do the job::
27
28  $ git diff --stat --summary vN-rc2..vN-rc3
29
30Here, there are two clear points in the history; Git will essentially
31"subtract" the beginning point from the end point and display the resulting
32differences.  The requested operation is unambiguous and easy enough to
33understand.
34
35When a subsystem maintainer creates a branch and commits changes to it, the
36result in the simplest case is a history that looks like::
37
38  ... vM --- vN-rc1 --- vN-rc2 --- vN-rc3 --- ... --- vN-rc7 --- vN
39                          |
40                          +-- c1 --- c2 --- ... --- cN
41
42If that maintainer now uses ``git diff`` to see what has changed between
43the mainline branch (let's call it "linus") and cN, there are still two
44clear endpoints, and the result is as expected.  So a pull request
45generated with ``git request-pull`` will also be as expected.  But now
46consider a slightly more complex development history::
47
48  ... vM --- vN-rc1 --- vN-rc2 --- vN-rc3 --- ... --- vN-rc7 --- vN
49                |         |
50                |         +-- c1 --- c2 --- ... --- cN
51                |                   /
52                +-- x1 --- x2 --- x3
53
54Our maintainer has created one branch at vN-rc1 and another at vN-rc2; the
55two were then subsequently merged into c2.  Now a pull request generated
56for cN may end up being messy indeed, and developers often end up wondering
57why.
58
59What is happening here is that there are no longer two clear end points for
60the ``git diff`` operation to use.  The development culminating in cN
61started in two different places; to generate the diffstat, ``git diff``
62ends up having pick one of them and hoping for the best.  If the diffstat
63starts at vN-rc1, it may end up including all of the changes between there
64and the second origin end point (vN-rc2), which is certainly not what our
65maintainer had in mind.  With all of that extra junk in the diffstat, it
66may be impossible to tell what actually happened in the changes leading up
67to cN.
68
69Maintainers often try to resolve this problem by, for example, rebasing the
70branch or performing another merge with the linus branch, then recreating
71the pull request.  This approach tends not to lead to joy at the receiving
72end of that pull request; rebasing and/or merging just before pushing
73upstream is a well-known way to get a grumpy response.
74
75So what is to be done?  The best response when confronted with this
76situation is to indeed to do a merge with the branch you intend your work
77to be pulled into, but to do it privately, as if it were the source of
78shame.  Create a new, throwaway branch and do the merge there::
79
80  ... vM --- vN-rc1 --- vN-rc2 --- vN-rc3 --- ... --- vN-rc7 --- vN
81                |         |                                      |
82                |         +-- c1 --- c2 --- ... --- cN           |
83                |                   /               |            |
84                +-- x1 --- x2 --- x3                +------------+-- TEMP
85
86The merge operation resolves all of the complications resulting from the
87multiple beginning points, yielding a coherent result that contains only
88the differences from the mainline branch.  Now it will be possible to
89generate a diffstat with the desired information::
90
91  $ git diff -C --stat --summary linus..TEMP
92
93Save the output from this command, then simply delete the TEMP branch;
94definitely do not expose it to the outside world.  Take the saved diffstat
95output and edit it into the messy pull request, yielding a result that
96shows what is really going on.  That request can then be sent upstream.
97