when merged more files are listed as changed

[Simon King]

On Thu, Jun 4, 2015 at 6:49 AM, Prabhakar S <prabhakar.s at tegile.com> wrote:
> On Wed, Jun 3, 2015 at 11:08 PM, Matt Mackall <mpm at selenic.com> wrote:
>>
>> On Wed, 2015-06-03 at 18:44 +0530, Prabhakar S wrote:
>> > Greetings,
>> >
>> > We have a scenario where for two releases we are having two repos and
>> > sometimes when the two repos are merged, say repo_REL1 is merged to
>> > repo_REL2 and for some conflicts with two or three files .
>> >
>> > Once they are resolved and pushed, the notification email from hg shows
>> > ,
>> > more number of files as changed than expected.
>> >
>> > When i checked that , looks like it is generated using command below
>> >
>> > hg diff -r <tip> -r <priortotip_notsure_ifitisexact prior_to_tip> |
>> > diffstat
>>
>> By default, it's going to show the difference between a commit and its
>> first parent.
>>
>> > Now the question is
>> >
>> > 1. Is the hg notification using the above approach to send the list of
>> > modified files ?
>>
>> Roughly.
>>
>> > 2. Why is it showing more files as modified though only few files are
>> > modified ?
>>
>> If, for instance, you merge your development branch into your release
>> branch with no conflicts.. the diff is still going to show _all your
>> development_ because all those files have changed.
>>
>
> Thank you for the details.
>
> When I checked that, yes according to the information above, when i run a
> diff between the tip revision and the first parten it shows lot files in the
> diffstat.
>
> What Iam not able to understand is, when I run 'hg diff -c <first_parent> ,
> then only one file shows up, however when the diff is run between the tip
> and the first_parent lot of files are shown up.

"hg diff -c <first_parent>" means "show me the diffs that were
introduced in <first_parent>". That has no relevance at all to the
diffs introduced by the merge.

Imagine this graph:

1-2-3-6
 \   /
  4-5

1 is the common ancestor. 2 and 3 were changes made on the first
branch, and 4 and 5 were made on the second branch. 6 is the merge
revision, so 3 is the first parent of the merge, and 5 is the second
parent. "hg diff -c 3" will show you the diffs between 2 and 3, which
is probably not what you were looking for.


>
> I understand that , if say 10 files are modified in a changeset_A and if it
> is merged to a release repo, and if no conflicts also, all those 10 files
> are going to be shown as modified.
>
> It would be great, if you give some good pointers to understand how this is
> handled.
>

Don't think of a revision as a set of changes - think of it as a new
snapshot of the entire state of the repository. When you use "hg
diff", you are asking for the differences between 2 snapshots. A merge
is just another snapshot.

The diffs introduced by a merge depend on your point of view. If you
were previously looking at the first parent, it would appear that the
merge had introduced all the changes from the second parent's branch.
However, from the second parent's point of view, it appears that all
the changes from the *first* parent's branch have suddenly appeared.

In terms of the graph above, the diffs from 3 to 6 would show all the
changes that had been merged in from 4 and 5. But the diffs from 5 to
6 would show all the changes from 2 and 3.

Does that help at all?

Simon