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.
$ hg init t
$ cd t
$ echo import > port
$ hg add port
$ hg commit -m 0 -u spam -d '0 0'
$ echo export >> port
$ hg commit -m 1 -u eggs -d '1 0'
$ echo export > port
$ echo vaportight >> port
$ echo 'import/export' >> port
$ hg commit -m 2 -u spam -d '2 0'
$ echo 'import/export' >> port
$ hg commit -m 3 -u eggs -d '3 0'
$ head -n 3 port > port1
$ mv port1 port
$ hg commit -m 4 -u spam -d '4 0'
pattern error
$ hg grep '**test**'
grep: invalid match pattern: nothing to repeat
[1]
simple
$ hg grep '.*'
port:4:export
port:4:vaportight
port:4:import/export
$ hg grep port port
port:4:export
port:4:vaportight
port:4:import/export
simple with color
$ hg --config extensions.color= grep --config color.mode=ansi \
> --color=always port port
\x1b[0;35mport\x1b[0m\x1b[0;36m:\x1b[0m\x1b[0;32m4\x1b[0m\x1b[0;36m:\x1b[0mex\x1b[0;31;1mport\x1b[0m (esc)
\x1b[0;35mport\x1b[0m\x1b[0;36m:\x1b[0m\x1b[0;32m4\x1b[0m\x1b[0;36m:\x1b[0mva\x1b[0;31;1mport\x1b[0might (esc)
\x1b[0;35mport\x1b[0m\x1b[0;36m:\x1b[0m\x1b[0;32m4\x1b[0m\x1b[0;36m:\x1b[0mim\x1b[0;31;1mport\x1b[0m/ex\x1b[0;31;1mport\x1b[0m (esc)
all
$ hg grep --traceback --all -nu port port
port:4:4:-:spam:import/export
port:3:4:+:eggs:import/export
port:2:1:-:spam:import
port:2:2:-:spam:export
port:2:1:+:spam:export
port:2:2:+:spam:vaportight
port:2:3:+:spam:import/export
port:1:2:+:eggs:export
port:0:1:+:spam:import
other
$ hg grep -l port port
port:4
$ hg grep import port
port:4:import/export
$ hg cp port port2
$ hg commit -m 4 -u spam -d '5 0'
follow
$ hg grep --traceback -f 'import\n\Z' port2
port:0:import
$ echo deport >> port2
$ hg commit -m 5 -u eggs -d '6 0'
$ hg grep -f --all -nu port port2
port2:6:4:+:eggs:deport
port:4:4:-:spam:import/export
port:3:4:+:eggs:import/export
port:2:1:-:spam:import
port:2:2:-:spam:export
port:2:1:+:spam:export
port:2:2:+:spam:vaportight
port:2:3:+:spam:import/export
port:1:2:+:eggs:export
port:0:1:+:spam:import
$ hg up -q null
$ hg grep -f port
[1]
$ cd ..
$ hg init t2
$ cd t2
$ hg grep foobar foo
[1]
$ hg grep foobar
[1]
$ echo blue >> color
$ echo black >> color
$ hg add color
$ hg ci -m 0
$ echo orange >> color
$ hg ci -m 1
$ echo black > color
$ hg ci -m 2
$ echo orange >> color
$ echo blue >> color
$ hg ci -m 3
$ hg grep orange
color:3:orange
$ hg grep --all orange
color:3:+:orange
color:2:-:orange
color:1:+:orange
match in last "line" without newline
$ $PYTHON -c 'fp = open("noeol", "wb"); fp.write("no infinite loop"); fp.close();'
$ hg ci -Amnoeol
adding noeol
$ hg grep loop
noeol:4:no infinite loop
$ cd ..
Issue685: traceback in grep -r after rename
Got a traceback when using grep on a single
revision with renamed files.
$ hg init issue685
$ cd issue685
$ echo octarine > color
$ hg ci -Amcolor
adding color
$ hg rename color colour
$ hg ci -Am rename
$ hg grep octarine
colour:1:octarine
color:0:octarine
Used to crash here
$ hg grep -r 1 octarine
colour:1:octarine
$ cd ..
Issue337: test that grep follows parent-child relationships instead
of just using revision numbers.
$ hg init issue337
$ cd issue337
$ echo white > color
$ hg commit -A -m "0 white"
adding color
$ echo red > color
$ hg commit -A -m "1 red"
$ hg update 0
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ echo black > color
$ hg commit -A -m "2 black"
created new head
$ hg update --clean 1
1 files updated, 0 files merged, 0 files removed, 0 files unresolved
$ echo blue > color
$ hg commit -A -m "3 blue"
$ hg grep --all red
color:3:-:red
color:1:+:red
$ cd ..
$ hg init a
$ cd a
$ cp "$TESTDIR/binfile.bin" .
$ hg add binfile.bin
$ hg ci -m 'add binfile.bin'
$ hg grep "MaCam" --all
binfile.bin:0:+: Binary file matches
$ cd ..