The home of 'Abort' is 'error' not 'util' however, a lot of code seems to be
confused about that and gives all the credit to 'util' instead of the
hardworking 'error'. In a spirit of equity, we break the cycle of injustice and
give back to 'error' the respect it deserves. And screw that 'util' poser.
For great justice.
Nodemap is not aware of filtering so we need to ask the changelog itself if a
node is known. This is probably a bit slower but such check does not dominated
discovery time. This is necessary if we want to run discovery on filtered repo.
There have been quite a few places where we pop elements off the
front of a list. This can turn O(n) algorithms into something more
like O(n**2). Python has provided a deque type that can do this
efficiently since at least 2.4.
As an example of the difference a deque can make, it improves
perfancestors performance on a Linux repo from 0.50 seconds to 0.36.
Old discovery only returned incoming heads, not all of them (for
changegroupsubset). New discovery must always return all of the remote heads
(for getbundle). I failed to properly adjust treediscovery in 43f4c1113c8d
when introducing setdiscovery.
The actual observable problem was 'remote: unsynced changes' when trying
to push a cset on one named branch to a server with a new cset on another
named branch. This scenario is now tested in test-treediscovery.t.
I ran the entire test suite with "known" and "getbundle" disabled in
localrepository. This generated failures because the old findoutgoing
had always queried remote's heads explicitly and thus always got them
back in the returned heads. treediscovery.findcommonincoming now
correctly returns remote's heads in all cases.
Also adds a dedicated test for running treediscovery against a
pre-getbundle HTTP server.
Adds a new discovery method based on repeatedly sampling the still
undecided subset of the local node graph to determine the set of nodes
common to both the client and the server.
For small differences between client and server, it uses about the same
or slightly fewer roundtrips than the old tree-based discovery. For
larger differences, it typically reduces the number of roundtrips
drastically (from 150 to 4, for instance).
The old discovery code now lives in treediscovery.py, the new code is
in setdiscovery.py.
Still missing is a hook for extensions to contribute nodes to the
initial sample. For instance, Augie's remotebranches could contribute
the last known state of the server's heads.
Credits for the actual sampler and computing common heads instead of
bases go to Benoit Boissinot.