I am working on a script that parses a mercurial log and generates a report of all changes between any two revisions. In order to handle branches, I believe I need to use revsets to generate ancestry of the final revision but I want to limit the scope to exclude any revisions which are ancestors to the starting revision.
10
/
/
9
|\
| \
| \
8 |
| 7
| 6 |
|/ |
5 4
| /
| /
|/
3
|
|
|
2
|
|
|
1
At 3, a feature branch is created from the default branch.
At 5, a release branch (6) is created from the default branch.
At 9, the feature branch is merged back to default and a new release (10) is created.
I want a list of all changes since 5 that affect 10 => 4,5,7,8,9,10
The difficulty I am having is limiting the ancestry search when 3 is reached via the feature branch. I don't want it since 3 was already a part of 5.
Other discussions I reviewed:
Close but didn't help with limiting the scope.
This one implies you know the branch but I need to find them all dynamically.
OK, if your business-task is "What's new in release", best iteration of revset will be ancestors(LAST) - ancestors(PREVIOUS), i.e. for your sample
hg log -r "ancestors(10) - ancestors(6)"
(note missing r5 in output, because it was in 6)
maybe revset will be more usable in [revsetalias] with alias added for better readability, smth.like (untested!!!)
[revsetalias]
new($1,$2) = ancestors($1) - ancestors($2)
[alias]
cl = log -r "new($1,$2)" --style changelog