diff mbox series

[1/3] docs: explain why squash merges are broken with long-running branches

Message ID 20200912204824.2824106-2-sandals@crustytoothpaste.net
State New
Headers show
Series FAQ entries for merges and modified files | expand

Commit Message

brian m. carlson Sept. 12, 2020, 8:48 p.m. UTC
In many projects, squash merges are commonly used, primarily to keep a
tidy history in the face of developers who do not use logically
independent, bisectable commits.  As common as this is, this tends to
cause significant problems when squash merges are used to merge
long-running branches due to the lack of any new merge bases.  Even very
experienced developers may make this mistakes, so let's add a FAQ entry
explaining why this is problematic and explaining that regular merge
commits should be used to merge two long-running branches.

Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
---
 Documentation/gitfaq.txt | 33 +++++++++++++++++++++++++++++++++
 1 file changed, 33 insertions(+)

Comments

Martin Ågren Sept. 13, 2020, 3:05 p.m. UTC | #1
On Sat, 12 Sep 2020 at 22:52, brian m. carlson
<sandals@crustytoothpaste.net> wrote:
>
> In many projects, squash merges are commonly used, primarily to keep a
> tidy history in the face of developers who do not use logically
> independent, bisectable commits.  As common as this is, this tends to
> cause significant problems when squash merges are used to merge
> long-running branches due to the lack of any new merge bases.  Even very
> experienced developers may make this mistakes, so let's add a FAQ entry

s/mistakes/mistake/

> explaining why this is problematic and explaining that regular merge
> commits should be used to merge two long-running branches.

> +What kinds of problems can occur when merging long-running branches with squash merges?::
> +       In general, there are a variety of problems that can occur when using squash
> +       merges with long-running branches.  These can include seeing extra commits in
> +       `git log` output, with a GUI, or when using the `...` notation to express a
> +       range, as well as the possibility of needing to re-resolve conflicts again and
> +       again.

Very well-written, as always.

I grepped around a little and couldn't find that we're using this "long-
running branch" notion anywhere. I think it might make sense to
define/explain that somewhere early, e.g., with something like "...
merged repeatedly ... more development...".

So in particular, a "long-running" branch is not something like

  I branched off a long time ago, I've been hacking on this feature for
  a long time and the branch has run quite long with lots of commits.
  Now it's time to merge it and be done with it!

Once that piece of background is in place, what you have here reads very
well to me. It was only somewhere halfway through the second paragraph
that I got what you meant by "long-running branch".


Martin
brian m. carlson Sept. 13, 2020, 5:12 p.m. UTC | #2
On 2020-09-13 at 15:05:31, Martin Ågren wrote:
> On Sat, 12 Sep 2020 at 22:52, brian m. carlson
> <sandals@crustytoothpaste.net> wrote:
> >
> > In many projects, squash merges are commonly used, primarily to keep a
> > tidy history in the face of developers who do not use logically
> > independent, bisectable commits.  As common as this is, this tends to
> > cause significant problems when squash merges are used to merge
> > long-running branches due to the lack of any new merge bases.  Even very
> > experienced developers may make this mistakes, so let's add a FAQ entry
> 
> s/mistakes/mistake/

Will fix.

> Very well-written, as always.
> 
> I grepped around a little and couldn't find that we're using this "long-
> running branch" notion anywhere. I think it might make sense to
> define/explain that somewhere early, e.g., with something like "...
> merged repeatedly ... more development...".
> 
> So in particular, a "long-running" branch is not something like
> 
>   I branched off a long time ago, I've been hacking on this feature for
>   a long time and the branch has run quite long with lots of commits.
>   Now it's time to merge it and be done with it!

Correct.  By "long-running branch", I mean a branch into which code
routinely gets integrated from feature branches.  For example, the
branches master and develop in a Git Flow workflow, or per-version
release branches if you use that workflow.  I didn't mean an otherwise
divergent branch which has persisted for a long time.

The problem happens, essentially, if you have two branches and you merge
one of them into another repeatedly with squash merges, so I'll be sure
to make that clear.  That will probably help with intelligibility for
non-native speakers as well.

> Once that piece of background is in place, what you have here reads very
> well to me. It was only somewhere halfway through the second paragraph
> that I got what you meant by "long-running branch".

Thanks, this is helpful feedback.
diff mbox series

Patch

diff --git a/Documentation/gitfaq.txt b/Documentation/gitfaq.txt
index 9cd7a592ac..550f4e30d6 100644
--- a/Documentation/gitfaq.txt
+++ b/Documentation/gitfaq.txt
@@ -241,6 +241,39 @@  How do I know if I want to do a fetch or a pull?::
 	ignore the upstream changes.  A pull consists of a fetch followed
 	immediately by either a merge or rebase.  See linkgit:git-pull[1].
 
+Merging and Rebasing
+--------------------
+
+[[long-running-squash-merge]]
+What kinds of problems can occur when merging long-running branches with squash merges?::
+	In general, there are a variety of problems that can occur when using squash
+	merges with long-running branches.  These can include seeing extra commits in
+	`git log` output, with a GUI, or when using the `...` notation to express a
+	range, as well as the possibility of needing to re-resolve conflicts again and
+	again.
++
+When Git does a normal merge between two branches, it considers exactly three
+points: the two branches and a third commit, called the _merge base_, which is
+usually the common ancestor of the commits.  The result of the merge is the sum
+of the changes between the merge base and each head.  When you merge two
+long-running branches with a regular merge commit, this results in a new commit
+which will end up as a merge base when they're merged again, because there is
+now a new common ancestor.  Git doesn't have to consider changes that occurred
+before the merge base, so you don't have to re-resolve any conflicts you
+resolved before.
++
+When you perform a squash merge, a merge commit isn't created; instead, the
+changes from one side are applied as a regular commit to the other side.  This
+means that the merge base for these branches won't have changed, and so when Git
+goes to perform its next merge, it considers all of the changes that it
+considered the last time plus the new changes.  That means any conflicts may
+need to be re-resolved.  Similarly, anything using the `...` notation in `git
+diff`, `git log`, or a GUI will result in showing all of the changes since the
+original merge base.
++
+As a consequence, if you want to merge two long-running branches, it's best to
+always use a regular merge commit.
+
 Hooks
 -----