[Feature] import rich history from Debian/upstream

[Christian Ehrhardt ]

Hi,
a session on git today made me think of the following feature and in a quick discussion did didn't sound too impossible.

We import debian uploads as a single commit.
But we might have a chance to do better.

In debian/sid:debian/control we can extract VCS info which we could fetch.
Also mnot all but many Debian packages follow the same tagging scheme (we might check a few if there are more common ones).
The procedure could be like:

If
- the treeish of tag salsa/debian/<pkg-version-old> matches debian/sid (old)
- the treeish of tag salsa/debian/<pkg-version-new> matches new synthesized content
- Then instead of committing the synthesized import cherry-pick all of salsa/debian/<pkg-version-old>..salsa/debian/<pkg-version-new>

The content should be the same, but we keep Debian commits individually which is better for
a) checking history
b) reverting or referencing individual commits
c) things like git tag --contains

Expected edge cases:
 - if the debian upload mismatches the git content (empty dir, d/control.in, ...) it won't match

Robie meant that git ubuntu is fairly modular and this could be just another hook identifying the potential candidates then passing it to the checkers we already have in place and it would work.
But that sounds too easy, so I'm convinced I skipped over the hard part of this - for sure I'd not know where to start right now :-)
But I'm convinced that as a feature this would help.

P.S. if a parse VCS and fetch there is implemented for this that would also server bug 1673710

[Christian Ehrhardt  (paelzer)]

FYI bug 1673710 seems to cover 90% of the benefit this bug here would bring.
And the other one seems easier to implement, so whenever you think "hmm can't fix 1941759, have a look at 1673710 at least" :-)