tests/test-unionrepo.t

changegroup: compute seen files as changesets are added (issue4750)

Before this patch, addchangegroup() would walk the changelog and compute
the set of seen files between applying changesets and applying
manifests. When cloning large repositories such as mozilla-central,
this consumed a non-trivial amount of time. On my MBP, this walk takes
~10s. On a dainty EC2 instance, this was measured to take ~125s! On the
latter machine, this delay was enough for the Mercurial server to
disconnect the client, thinking it had timed out, thus causing a clone
to abort.

This patch enables the changelog to compute the set of changed files as
new revisions are added. By doing so, we:

* avoid a potentially heavy computation between changelog and manifest
  processing by spreading the computation across all changelog additions
* avoid extra reads from the changelog by operating on the data as it is
  added

The downside of this is that the add revision callback does result in
extra I/O. Before, we would perform a flush (and subsequent read to
construct the full revision) when new delta chains were created. For
changelogs, this is typically every 2-4 revisions. Using the callback
guarantees there will be a flush after every added revision *and* an
open + read of the changelog to obtain the full revision in order to
read the added files. So, this increases the frequency of these
operations by the average chain length. In the future, the revlog
should be smart enough to know how to read revisions that haven't been
flushed yet, thus eliminating this extra I/O.

On my MBP, the total CPU times for an `hg unbundle` with a local
mozilla-central gzip bundle containing 251,934 changesets and 211,065
files did not have a statistically significant change with this patch,
holding steady around 360s. So, the increased revlog flushing did not
have an effect.

With this patch, there is no longer a visible pause between applying
changeset and manifest data. Before, it sure felt like Mercurial was
lethargic making this transition. Now, the transition is nearly
instantaneous, giving the impression that Mercurial is faster. Of course,
eliminating this pause means that the potential for network disconnect due
to channel inactivity during the changelog walk is eliminated as well.
And that is the impetus behind this change.

Test unionrepo functionality

Create one repository

  $ hg init repo1
  $ cd repo1
  $ touch repo1-0
  $ echo repo1-0 > f
  $ hg ci -Aqmrepo1-0
  $ touch repo1-1
  $ echo repo1-1 >> f
  $ hg ci -Aqmrepo1-1
  $ touch repo1-2
  $ echo repo1-2 >> f
  $ hg ci -Aqmrepo1-2
  $ hg log --template '{rev}:{node|short}  {desc|firstline}\n'
  2:68c0685446a3  repo1-2
  1:8a58db72e69d  repo1-1
  0:f093fec0529b  repo1-0
  $ tip1=`hg id -q`
  $ cd ..

- and a clone with a not-completely-trivial history

  $ hg clone -q repo1 --rev 0 repo2
  $ cd repo2
  $ touch repo2-1
  $ sed '1i\
  > repo2-1 at top
  > ' f > f.tmp
  $ mv f.tmp f
  $ hg ci -Aqmrepo2-1
  $ touch repo2-2
  $ hg pull -q ../repo1 -r 1
  $ hg merge -q
  $ hg ci -Aqmrepo2-2-merge
  $ touch repo2-3
  $ echo repo2-3 >> f
  $ hg ci -mrepo2-3
  $ hg log --template '{rev}:{node|short}  {desc|firstline}\n'
  4:2f0d178c469c  repo2-3
  3:9e6fb3e0b9da  repo2-2-merge
  2:8a58db72e69d  repo1-1
  1:c337dba826e7  repo2-1
  0:f093fec0529b  repo1-0
  $ cd ..

revisions from repo2 appear as appended / pulled to repo1

  $ hg -R union:repo1+repo2 log --template '{rev}:{node|short}  {desc|firstline}\n'
  5:2f0d178c469c  repo2-3
  4:9e6fb3e0b9da  repo2-2-merge
  3:c337dba826e7  repo2-1
  2:68c0685446a3  repo1-2
  1:8a58db72e69d  repo1-1
  0:f093fec0529b  repo1-0

manifest can be retrieved for revisions in both repos

  $ hg -R union:repo1+repo2 mani -r $tip1
  f
  repo1-0
  repo1-1
  repo1-2
  $ hg -R union:repo1+repo2 mani -r 4
  f
  repo1-0
  repo1-1
  repo2-1
  repo2-2

files can be retrieved form both repos

  $ hg -R repo1 cat repo1/f -r2
  repo1-0
  repo1-1
  repo1-2

  $ hg -R union:repo1+repo2 cat -r$tip1 repo1/f
  repo1-0
  repo1-1
  repo1-2

  $ hg -R union:repo1+repo2 cat -r4 $TESTTMP/repo1/f
  repo2-1 at top
  repo1-0
  repo1-1

files can be compared across repos

  $ hg -R union:repo1+repo2 diff -r$tip1 -rtip
  diff -r 68c0685446a3 -r 2f0d178c469c f
  --- a/f	Thu Jan 01 00:00:00 1970 +0000
  +++ b/f	Thu Jan 01 00:00:00 1970 +0000
  @@ -1,3 +1,4 @@
  +repo2-1 at top
   repo1-0
   repo1-1
  -repo1-2
  +repo2-3

heads from both repos are found correctly

  $ hg -R union:repo1+repo2 heads --template '{rev}:{node|short}  {desc|firstline}\n'
  5:2f0d178c469c  repo2-3
  2:68c0685446a3  repo1-2

revsets works across repos

  $ hg -R union:repo1+repo2 id -r "ancestor($tip1, 5)"
  8a58db72e69d

annotate works - an indication that linkrevs works

  $ hg --cwd repo1 -Runion:../repo2 annotate $TESTTMP/repo1/f -r tip
  3: repo2-1 at top
  0: repo1-0
  1: repo1-1
  5: repo2-3

union repos can be cloned ... and clones works correctly

  $ hg clone -U union:repo1+repo2 repo3
  requesting all changes
  adding changesets
  adding manifests
  adding file changes
  added 6 changesets with 11 changes to 6 files (+1 heads)

  $ hg -R repo3 paths
  default = union:repo1+repo2

  $ hg -R repo3 verify
  checking changesets
  checking manifests
  crosschecking files in changesets and manifests
  checking files
  6 files, 6 changesets, 11 total revisions

  $ hg -R repo3 heads --template '{rev}:{node|short}  {desc|firstline}\n'
  5:2f0d178c469c  repo2-3
  2:68c0685446a3  repo1-2

  $ hg -R repo3 log --template '{rev}:{node|short}  {desc|firstline}\n'
  5:2f0d178c469c  repo2-3
  4:9e6fb3e0b9da  repo2-2-merge
  3:c337dba826e7  repo2-1
  2:68c0685446a3  repo1-2
  1:8a58db72e69d  repo1-1
  0:f093fec0529b  repo1-0