[MERGE] Add --no-pending to status to not show the pending merges. (#202830)

[John Arbash Meinel]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Robert Collins wrote:
> On Sun, 2008-03-16 at 15:23 +0000, James Westby wrote:
>> Hi,
>>
>> It was requested that there be an option to turn off the display
>> of pending merges, as doing that is slow on large histories. This
>> patch does that.
>>
>> It is just showing the summaries etc. that is slow, so it is possible
>> that we could give some notification that there are pending merges 
>> still. Does anyone have an opinion on doing that? Even showing just
>> the summaries of the tips of the merged branches would be pretty cheap,
>> it's showing the rest of the revisions that you will be getting that
>> is the slowest part.
> 
> I think its good to address user issues rapidly; but we should at least
> attempt to do the correct long term fix before adding workarounds;
> workarounds only fix the issue for some users and are really 'unbreakme'
> options.
> 
> I think there are two things here.
> 1) Showing many revisions on huge merges will be very clumsy.
> 2) Its slower than it should be (duh).
> 
> For 1, I think we should file a bug for ui improvements.
> For 2, please get a --lsprof-file trace and we can probably identify the
> issue very rapidly.
> 
> -Rob

We already have a bug open on it (status after merge with packs).

And the bug is because we do a "set(ancestry1) - set(ancestry2)" stuff
to figure out what should be showed.

We have a function Graph.find_difference() which computes that by
walking the graph, but it can (ATM) return too many nodes because it
doesn't walk far enough on the "common" nodes. (I have some test cases,
and we have some XFAIL ones already merged into bzr.dev)

The decision at the time was that we can't use find_difference() until
it is correct. I think I made it correct, but it didn't seem
particularly fast, and then it seemed like I had spent too much time
already without getting better results. (I think we ended up needing me
to work on annotate performance, etc.)

I still have the branch.

John
=:->
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFH3WjhJdeBCYSNAAMRAr4sAJ0YNmx+WWV4Z78oxWFw57cRk1w+zACgnvpD
CyQIJbsUdZKzTdETNOetLCg=
=I2/z
-----END PGP SIGNATURE-----