Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 14 Mar 2022 05:59:20 +0100] rev 49082
discovery: also audit the number of queries done
In addition to the number of roundtrip, we now also track the number of queries
we perform, this is useful to assert the tradeoff between number of roundtrip and
the number of queries.
Differential Revision: https://phab.mercurial-scm.org/D12398
Pierre-Yves David <pierre-yves.david@octobus.net> [Sun, 13 Mar 2022 16:10:53 +0100] rev 49081
search-discovery-case: display more information about the interresting case
We display information about the total number of revs and the common/missing
numbers. This is useful to spot the interresting case.
Differential Revision: https://phab.mercurial-scm.org/D12397
Pierre-Yves David <pierre-yves.david@octobus.net> [Mon, 21 Mar 2022 20:06:51 +0100] rev 49080
subsetmaker: rework the antichain generation to be usable
Before this, antichain computation can run for 10s of hours without completion in
sight. We use a more direct approach in the computation to keep the computation
in complexity in check. With good result.
We can now have a full antichain computation on mozilla-try in about one
minute. Which is usable.
Differential Revision: https://phab.mercurial-scm.org/D12396
Pierre-Yves David <pierre-yves.david@octobus.net> [Sun, 13 Mar 2022 16:24:01 +0100] rev 49079
subsetmaker: use SortedSet for the scratch variant
This provides a massive speedup on wide repository with many heads. For example
on mozilla-try, this move from un-usable slow to fairly instant.
Differential Revision: https://phab.mercurial-scm.org/D12395
Pierre-Yves David <pierre-yves.david@octobus.net> [Sun, 13 Mar 2022 15:53:29 +0100] rev 49078
subsetmaker: stabilize the computation of `scratch` subset
`heads` is set, order of the element are not deterministic and we need to
stabilize that if we want to get reproducible results.
Differential Revision: https://phab.mercurial-scm.org/D12394
Joerg Sonnenberger <joerg@bec.de> [Tue, 20 Jul 2021 15:07:10 +0200] rev 49077
revlog: recommit 49fd21f32695 with a fix for issue6528
`filelog.size` currently special cases two forms of metadata encoding:
- copy data via the parent order as flag bit
- censor data by peaking into the raw delta
All other forms of metadata encoding including the empty metadata block
are mishandled. In `basefilectx.cmp` the empty metadata block is
explicitly checked to compensate for this.
Restore 49fd21f32695, but disable it for filelog, so that the original
flag bit use contines to work. Document all this mess for now in
preparation of a proper rework.
Differential Revision: https://phab.mercurial-scm.org/D11203