tests/test-dirstate-race.t

setdiscovery: stop limiting the number of local head we initially send

In our testing this limitation provides now real gain and instead triggers
pathological discovery timing for some repository with many heads.

See inline documentation for details.

Some timing below:

Mozilla try repository, (~1M revs, ~35K heads), discovery between 2 clones with
100 head missing on each side

before:
! wall 1.492111 comb 1.490000 user 1.450000 sys 0.040000 (best of 20)
! wall 1.813992 comb 1.820000 user 1.700000 sys 0.120000 (max of 20)
! wall 1.574326 comb 1.573500 user 1.522000 sys 0.051500 (avg of 20)
! wall 1.572583 comb 1.570000 user 1.520000 sys 0.050000 (median of 20)

after:
! wall 1.147834 comb 1.150000 user 1.090000 sys 0.060000 (best of 20)
! wall 1.449144 comb 1.450000 user 1.330000 sys 0.120000 (max of 20)
! wall 1.204618 comb 1.202500 user 1.146500 sys 0.056000 (avg of 20)
! wall 1.194407 comb 1.190000 user 1.140000 sys 0.050000 (median of 20)


pypy (~100 heads, 317 heads) discovery between clones with only 42 common heads

before:
! wall 0.031653 comb 0.030000 user 0.030000 sys 0.000000 (best of 25)
! wall 0.055719 comb 0.050000 user 0.040000 sys 0.010000 (max of 25)
! wall 0.038939 comb 0.039600 user 0.038400 sys 0.001200 (avg of 25)
! wall 0.038660 comb 0.050000 user 0.040000 sys 0.010000 (median of 25)

after:
! wall 0.018754 comb 0.020000 user 0.020000 sys 0.000000 (best of 49)
! wall 0.034505 comb 0.040000 user 0.030000 sys 0.010000 (max of 49)
! wall 0.019631 comb 0.019796 user 0.018367 sys 0.001429 (avg of 49)
! wall 0.019132 comb 0.020000 user 0.020000 sys 0.000000 (median of 49)


Private repository (~1M revs, ~3K heads), discovery from a strip subset, about
100 changesets to be pulled.

before:
! wall 1.837729 comb 1.840000 user 1.790000 sys 0.050000 (best of 20)
! wall 2.203468 comb 2.200000 user 2.100000 sys 0.100000 (max of 20)
! wall 2.049355 comb 2.048500 user 2.002500 sys 0.046000 (avg of 20)
! wall 2.035315 comb 2.040000 user 2.000000 sys 0.040000 (median of 20)

after:
! wall 0.136598 comb 0.130000 user 0.110000 sys 0.020000 (best of 20)
! wall 0.330519 comb 0.330000 user 0.260000 sys 0.070000 (max of 20)
! wall 0.157254 comb 0.155500 user 0.123000 sys 0.032500 (avg of 20)
! wall 0.149870 comb 0.140000 user 0.110000 sys 0.030000 (median of 20)


Same private repo, discovery between two clone with 500 different heads on each
side:

before:
! wall 2.372919 comb 2.370000 user 2.320000 sys 0.050000 (best of 20)
! wall 2.622422 comb 2.610000 user 2.510000 sys 0.100000 (max of 20)
! wall 2.450135 comb 2.450000 user 2.402000 sys 0.048000 (avg of 20)
! wall 2.443896 comb 2.450000 user 2.410000 sys 0.040000 (median of 20)

after:
! wall 0.625497 comb 0.620000 user 0.570000 sys 0.050000 (best of 20)
! wall 0.834723 comb 0.820000 user 0.730000 sys 0.090000 (max of 20)
! wall 0.675725 comb 0.675500 user 0.628000 sys 0.047500 (avg of 20)
! wall 0.671614 comb 0.680000 user 0.640000 sys 0.040000 (median of 20)

  $ hg init repo
  $ cd repo
  $ echo a > a
  $ hg add a
  $ hg commit -m test

Do we ever miss a sub-second change?:

  $ for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20; do
  >     hg co -qC 0
  >     echo b > a
  >     hg st
  > done
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a
  M a

  $ echo test > b
  $ mkdir dir1
  $ echo test > dir1/c
  $ echo test > d

  $ echo test > e
#if execbit
A directory will typically have the execute bit -- make sure it doesn't get
confused with a file with the exec bit set
  $ chmod +x e
#endif

  $ hg add b dir1 d e
  adding dir1/c
  $ hg commit -m test2

  $ cat >> $TESTTMP/dirstaterace.py << EOF
  > from mercurial import (
  >     context,
  >     extensions,
  > )
  > def extsetup(ui):
  >     extensions.wrapfunction(context.workingctx, '_checklookup', overridechecklookup)
  > def overridechecklookup(orig, self, files):
  >     # make an update that changes the dirstate from underneath
  >     self._repo.ui.system(br"sh '$TESTTMP/dirstaterace.sh'",
  >                          cwd=self._repo.root)
  >     return orig(self, files)
  > EOF

  $ hg debugrebuilddirstate
  $ hg debugdirstate
  n   0         -1 unset               a
  n   0         -1 unset               b
  n   0         -1 unset               d
  n   0         -1 unset               dir1/c
  n   0         -1 unset               e

XXX Note that this returns M for files that got replaced by directories. This is
definitely a bug, but the fix for that is hard and the next status run is fine
anyway.

  $ cat > $TESTTMP/dirstaterace.sh <<EOF
  > rm b && rm -r dir1 && rm d && mkdir d && rm e && mkdir e
  > EOF

  $ hg status --config extensions.dirstaterace=$TESTTMP/dirstaterace.py
  M d
  M e
  ! b
  ! dir1/c
  $ hg debugdirstate
  n 644          2 * a (glob)
  n   0         -1 unset               b
  n   0         -1 unset               d
  n   0         -1 unset               dir1/c
  n   0         -1 unset               e

  $ hg status
  ! b
  ! d
  ! dir1/c
  ! e

  $ rmdir d e
  $ hg update -C -q .

Test that dirstate changes aren't written out at the end of "hg
status", if .hg/dirstate is already changed simultaneously before
acquisition of wlock in workingctx._poststatusfixup().

This avoidance is important to keep consistency of dirstate in race
condition (see issue5584 for detail).

  $ hg parents -q
  1:* (glob)

  $ hg debugrebuilddirstate
  $ hg debugdirstate
  n   0         -1 unset               a
  n   0         -1 unset               b
  n   0         -1 unset               d
  n   0         -1 unset               dir1/c
  n   0         -1 unset               e

  $ cat > $TESTTMP/dirstaterace.sh <<EOF
  > # This script assumes timetable of typical issue5584 case below:
  > #
  > # 1. "hg status" loads .hg/dirstate
  > # 2. "hg status" confirms clean-ness of FILE
  > # 3. "hg update -C 0" updates the working directory simultaneously
  > #    (FILE is removed, and FILE is dropped from .hg/dirstate)
  > # 4. "hg status" acquires wlock
  > #    (.hg/dirstate is re-loaded = no FILE entry in dirstate)
  > # 5. "hg status" marks FILE in dirstate as clean
  > #    (FILE entry is added to in-memory dirstate)
  > # 6. "hg status" writes dirstate changes into .hg/dirstate
  > #    (FILE entry is written into .hg/dirstate)
  > #
  > # To reproduce similar situation easily and certainly, #2 and #3
  > # are swapped.  "hg cat" below ensures #2 on "hg status" side.
  > 
  > hg update -q -C 0
  > hg cat -r 1 b > b
  > EOF

"hg status" below should excludes "e", of which exec flag is set, for
portability of test scenario, because unsure but missing "e" is
treated differently in _checklookup() according to runtime platform.

- "missing(!)" on POSIX, "pctx[f].cmp(self[f])" raises ENOENT
- "modified(M)" on Windows, "self.flags(f) != pctx.flags(f)" is True

  $ hg status --config extensions.dirstaterace=$TESTTMP/dirstaterace.py --debug -X path:e
  skip updating dirstate: identity mismatch
  M a
  ! d
  ! dir1/c

  $ hg parents -q
  0:* (glob)
  $ hg files
  a
  $ hg debugdirstate
  n * * * a (glob)

  $ rm b

#if fsmonitor

Create fsmonitor state.

  $ hg status
  $ f --type .hg/fsmonitor.state
  .hg/fsmonitor.state: file

Test that invalidating fsmonitor state in the middle (which doesn't require the
wlock) causes the fsmonitor update to be skipped.
hg debugrebuilddirstate ensures that the dirstaterace hook will be called, but
it also invalidates the fsmonitor state. So back it up and restore it.

  $ mv .hg/fsmonitor.state .hg/fsmonitor.state.tmp
  $ hg debugrebuilddirstate
  $ mv .hg/fsmonitor.state.tmp .hg/fsmonitor.state

  $ cat > $TESTTMP/dirstaterace.sh <<EOF
  > rm .hg/fsmonitor.state
  > EOF

  $ hg status --config extensions.dirstaterace=$TESTTMP/dirstaterace.py --debug
  skip updating fsmonitor.state: identity mismatch
  $ f .hg/fsmonitor.state
  .hg/fsmonitor.state: file not found

#endif

Set up a rebase situation for issue5581.

  $ echo c2 > a
  $ echo c2 > b
  $ hg add b
  $ hg commit -m c2
  created new head
  $ echo c3 >> a
  $ hg commit -m c3
  $ hg update 2
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ echo c4 >> a
  $ echo c4 >> b
  $ hg commit -m c4
  created new head

Configure a merge tool that runs status in the middle of the rebase. The goal of
the status call is to trigger a potential bug if fsmonitor's state is written
even though the wlock is held by another process. The output of 'hg status' in
the merge tool goes to /dev/null because we're more interested in the results of
'hg status' run after the rebase.

  $ cat >> $TESTTMP/mergetool-race.sh << EOF
  > echo "custom merge tool"
  > printf "c2\nc3\nc4\n" > \$1
  > hg --cwd "$TESTTMP/repo" status > /dev/null
  > echo "custom merge tool end"
  > EOF
  $ cat >> $HGRCPATH << EOF
  > [extensions]
  > rebase =
  > [merge-tools]
  > test.executable=sh
  > test.args=$TESTTMP/mergetool-race.sh \$output
  > EOF

  $ hg rebase -s . -d 3 --tool test
  rebasing 4:b08445fd6b2a "c4" (tip)
  merging a
  custom merge tool
  custom merge tool end
  saved backup bundle to $TESTTMP/repo/.hg/strip-backup/* (glob)

This hg status should be empty, whether or not fsmonitor is enabled (issue5581).

  $ hg status