view tests/test-update-branches.t @ 26379:39d643252b9f

revlog: use existing file handle when reading during _addrevision

_addrevision() may need to read from revlogs as part of computing
deltas. Previously, we would flush existing file handles and open
a new, short-lived file handle to perform the reading.

If we have an existing file handle, it seems logical to reuse it
for reading instead of opening a new file handle. This patch
makes that the new behavior.

After this patch, revlog files are only reopened when adding
revisions if the revlog is switched from inline to non-inline.

On Linux when unbundling a bundle of the mozilla-central repo, this
patch has the following impact on system call counts:

Call     Before     After       Delta
write    827,639    673,390   -154,249
open     700,103    684,089    -16,014
read      74,489     74,489          0
fstat    493,924    461,896    -32,028
close    249,131    233,117    -16,014
stat     242,001    242,001          0
lstat     18,676     18,676          0
lseek     20,268     20,268          0
ioctl     14,652     13,173     -1,479
TOTAL  3,180,758  2,930,679   -250,079

It's worth noting that many of the open() calls fail due to missing
files. That's why there are many more open() calls than close().

Despite the significant system call reduction, this change does not
seem to have a significant performance impact on Linux.

On Windows 10 (not a VM, on a SSD), this patch appears to reduce
unbundle time for mozilla-central from ~960s to ~920s. This isn't
as significant as I was hoping. But a decrease it is nonetheless.
Still, Windows unbundle performance is still >2x slower than Linux.

Despite the lack of significant gains, fewer system calls is fewer
system calls. If nothing else, this will narrow the focus of potential
areas to optimize in the future.

# Construct the following history tree:
#
# @  5:e1bb631146ca  b1
# |
# o  4:a4fdb3b883c4 0:b608b9236435  b1
# |
# | o  3:4b57d2520816 1:44592833ba9f
# | |
# | | o  2:063f31070f65
# | |/
# | o  1:44592833ba9f
# |/
# o  0:b608b9236435

  $ mkdir b1
  $ cd b1
  $ hg init
  $ echo foo > foo
  $ echo zero > a
  $ hg init sub
  $ echo suba > sub/suba
  $ hg --cwd sub ci -Am addsuba
  adding suba
  $ echo 'sub = sub' > .hgsub
  $ hg ci -qAm0
  $ echo one > a ; hg ci -m1
  $ echo two > a ; hg ci -m2
  $ hg up -q 1
  $ echo three > a ; hg ci -qm3
  $ hg up -q 0
  $ hg branch -q b1
  $ echo four > a ; hg ci -qm4
  $ echo five > a ; hg ci -qm5

Initial repo state:

  $ hg log -G --template '{rev}:{node|short} {parents} {branches}\n'
  @  5:ff252e8273df  b1
  |
  o  4:d047485b3896 0:60829823a42a  b1
  |
  | o  3:6efa171f091b 1:0786582aa4b1
  | |
  | | o  2:bd10386d478c
  | |/
  | o  1:0786582aa4b1
  |/
  o  0:60829823a42a
  

Make sure update doesn't assume b1 is a repository if invoked from outside:

  $ cd ..
  $ hg update b1
  abort: no repository found in '$TESTTMP' (.hg not found)!
  [255]
  $ cd b1

Test helper functions:

  $ revtest () {
  >     msg=$1
  >     dirtyflag=$2   # 'clean', 'dirty' or 'dirtysub'
  >     startrev=$3
  >     targetrev=$4
  >     opt=$5
  >     hg up -qC $startrev
  >     test $dirtyflag = dirty && echo dirty > foo
  >     test $dirtyflag = dirtysub && echo dirty > sub/suba
  >     hg up $opt $targetrev
  >     hg parent --template 'parent={rev}\n'
  >     hg stat -S
  > }

  $ norevtest () {
  >     msg=$1
  >     dirtyflag=$2   # 'clean', 'dirty' or 'dirtysub'
  >     startrev=$3
  >     opt=$4
  >     hg up -qC $startrev
  >     test $dirtyflag = dirty && echo dirty > foo
  >     test $dirtyflag = dirtysub && echo dirty > sub/suba
  >     hg up $opt
  >     hg parent --template 'parent={rev}\n'
  >     hg stat -S
  > }

Test cases are documented in a table in the update function of merge.py.
Cases are run as shown in that table, row by row.

  $ norevtest 'none clean linear' clean 4
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  parent=5

  $ norevtest 'none clean same'   clean 2
  abort: not a linear update
  (merge or update --check to force update)
  parent=2


  $ revtest 'none clean linear' clean 1 2
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  parent=2

  $ revtest 'none clean same'   clean 2 3
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  parent=3

  $ revtest 'none clean cross'  clean 3 4
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  parent=4


  $ revtest 'none dirty linear' dirty 1 2
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  parent=2
  M foo

  $ revtest 'none dirtysub linear' dirtysub 1 2
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  parent=2
  M sub/suba

  $ revtest 'none dirty same'   dirty 2 3
  abort: uncommitted changes
  (commit or update --clean to discard changes)
  parent=2
  M foo

  $ revtest 'none dirtysub same'   dirtysub 2 3
  abort: uncommitted changes
  (commit or update --clean to discard changes)
  parent=2
  M sub/suba

  $ revtest 'none dirty cross'  dirty 3 4
  abort: uncommitted changes
  (commit or update --clean to discard changes)
  parent=3
  M foo

  $ norevtest 'none dirty cross'  dirty 2
  abort: uncommitted changes
  (commit and merge, or update --clean to discard changes)
  parent=2
  M foo

  $ revtest 'none dirtysub cross'  dirtysub 3 4
  abort: uncommitted changes
  (commit or update --clean to discard changes)
  parent=3
  M sub/suba

  $ revtest '-C dirty linear'   dirty 1 2 -C
  2 files updated, 0 files merged, 0 files removed, 0 files unresolved
  parent=2

  $ revtest '-c dirty linear'   dirty 1 2 -c
  abort: uncommitted changes
  parent=1
  M foo

  $ revtest '-c dirtysub linear'   dirtysub 1 2 -c
  abort: uncommitted changes in subrepository 'sub'
  parent=1
  M sub/suba

  $ norevtest '-c clean same'   clean 2 -c
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  parent=3

  $ revtest '-cC dirty linear'  dirty 1 2 -cC
  abort: cannot specify both -c/--check and -C/--clean
  parent=1
  M foo

Test obsolescence behavior
---------------------------------------------------------------------

successors should be taken in account when checking head destination

  $ cat << EOF >> $HGRCPATH
  > [ui]
  > logtemplate={rev}:{node|short} {desc|firstline}
  > [experimental]
  > evolution=createmarkers
  > EOF

Test no-argument update to a successor of an obsoleted changeset

  $ hg log -G
  o  5:ff252e8273df 5
  |
  o  4:d047485b3896 4
  |
  | o  3:6efa171f091b 3
  | |
  | | o  2:bd10386d478c 2
  | |/
  | @  1:0786582aa4b1 1
  |/
  o  0:60829823a42a 0
  
  $ hg book bm -r 3
  $ hg status
  M foo

We add simple obsolescence marker between 3 and 4 (indirect successors)

  $ hg id --debug -i -r 3
  6efa171f091b00a3c35edc15d48c52a498929953
  $ hg id --debug -i -r 4
  d047485b3896813b2a624e86201983520f003206
  $ hg debugobsolete 6efa171f091b00a3c35edc15d48c52a498929953 aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa
  $ hg debugobsolete aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa d047485b3896813b2a624e86201983520f003206

Test that 5 is detected as a valid destination from 3 and also accepts moving
the bookmark (issue4015)

  $ hg up --quiet --hidden 3
  $ hg up 5
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  $ hg book bm
  moving bookmark 'bm' forward from 6efa171f091b
  $ hg bookmarks
   * bm                        5:ff252e8273df

Test that 4 is detected as the no-argument destination from 3 and also moves
the bookmark with it
  $ hg up --quiet 0          # we should be able to update to 3 directly
  $ hg up --quiet --hidden 3 # but not implemented yet.
  $ hg book -f bm
  $ hg up
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved
  updating bookmark bm
  $ hg book
   * bm                        4:d047485b3896

Test that 5 is detected as a valid destination from 1
  $ hg up --quiet 0          # we should be able to update to 3 directly
  $ hg up --quiet --hidden 3 # but not implemented yet.
  $ hg up 5
  1 files updated, 0 files merged, 0 files removed, 0 files unresolved

Test that 5 is not detected as a valid destination from 2
  $ hg up --quiet 0
  $ hg up --quiet 2
  $ hg up 5
  abort: uncommitted changes
  (commit or update --clean to discard changes)
  [255]

Test that we don't crash when updating from a pruned changeset (i.e. has no
successors). Behavior should probably be that we update to the first
non-obsolete parent but that will be decided later.
  $ hg id --debug -r 2
  bd10386d478cd5a9faf2e604114c8e6da62d3889
  $ hg up --quiet 0
  $ hg up --quiet 2
  $ hg debugobsolete bd10386d478cd5a9faf2e604114c8e6da62d3889
  $ hg up
  0 files updated, 0 files merged, 0 files removed, 0 files unresolved