view tests/test-audit-path.t @ 40021:c537144fdbef

wireprotov2: support response caching

One of the things I've learned from managing VCS servers over the
years is that they are hard to scale. It is well known that some
companies have very beefy (read: very expensive) servers to power
their VCS needs. It is also known that specialized servers for
various VCS exist in order to facilitate scaling servers. (Mercurial
is in this boat.)

One of the aspects that make a VCS server hard to scale is the
high CPU load incurred by constant client clone/pull operations.

To alleviate the scaling pain associated with data retrieval
operations, I want to integrate caching into the Mercurial wire
protocol server as robustly as possible such that servers can
aggressively cache responses and defer as much server load as
possible.

This commit represents the initial implementation of a general
caching layer in wire protocol version 2.

We define a new interface and behavior for a wire protocol cacher
in repository.py. (This is probably where a reviewer should look
first to understand what is going on.)

The bulk of the added code is in wireprotov2server.py, where we
define how a command can opt in to being cached and integrate
caching into command dispatching.

From a very high-level:

* A command can declare itself as cacheable by providing a callable
  that can be used to derive a cache key.
* At dispatch time, if a command is cacheable, we attempt to
  construct a cacher and use it for serving the request and/or
  caching the request.
* The dispatch layer handles the bulk of the business logic for
  caching, making cachers mostly "dumb content stores."
* The mechanism for invalidating cached entries (one of the harder
  parts about caching in general) is by varying the cache key when
  state changes. As such, cachers don't need to be concerned with
  cache invalidation.

Initially, we've hooked up support for caching "manifestdata" and
"filedata" commands. These are the simplest to cache, as they should
be immutable over time. Caching of commands related to changeset
data is a bit harder (because cache validation is impacted by
changes to bookmarks, phases, etc). This will be implemented later.
(Strictly speaking, censoring a file should invalidate caches. I've
added an inline TODO to track this edge case.)

To prove it works, this commit implements a test-only extension
providing in-memory caching backed by an lrucachedict. A new test
showing this extension behaving properly is added. FWIW, the
cacher is ~50 lines of code, demonstrating the relative ease with
which a cache can be added to a server.

While the test cacher is not suitable for production workloads, just
for kicks I performed a clone of just the changeset and manifest data
for the mozilla-unified repository. With a fully warmed cache (of just
the manifest data since changeset data is not cached), server-side
CPU usage dropped from ~73s to ~28s. That's pretty significant and
demonstrates the potential that response caching has on server
scalability!

Differential Revision: https://phab.mercurial-scm.org/D4773

The simple store doesn't escape paths robustly and can't store paths
with periods, etc. So much of this test fails with it.
#require no-reposimplestore

  $ hg init

audit of .hg

  $ hg add .hg/00changelog.i
  abort: path contains illegal component: .hg/00changelog.i
  [255]

#if symlink

Symlinks

  $ mkdir a
  $ echo a > a/a
  $ hg ci -Ama
  adding a/a
  $ ln -s a b
  $ echo b > a/b
  $ hg add b/b
  abort: path 'b/b' traverses symbolic link 'b'
  [255]
  $ hg add b

should still fail - maybe

  $ hg add b/b
  abort: path 'b/b' traverses symbolic link 'b'
  [255]

  $ hg commit -m 'add symlink b'


Test symlink traversing when accessing history:
-----------------------------------------------

(build a changeset where the path exists as a directory)

  $ hg up 0
  0 files updated, 0 files merged, 1 files removed, 0 files unresolved
  $ mkdir b
  $ echo c > b/a
  $ hg add b/a
  $ hg ci -m 'add directory b'
  created new head

Test that hg cat does not do anything wrong the working copy has 'b' as directory

  $ hg cat b/a
  c
  $ hg cat -r "desc(directory)" b/a
  c
  $ hg cat -r "desc(symlink)" b/a
  b/a: no such file in rev bc151a1f53bd
  [1]

Test that hg cat does not do anything wrong the working copy has 'b' as a symlink (issue4749)

  $ hg up 'desc(symlink)'
  1 files updated, 0 files merged, 1 files removed, 0 files unresolved
  $ hg cat b/a
  b/a: no such file in rev bc151a1f53bd
  [1]
  $ hg cat -r "desc(directory)" b/a
  c
  $ hg cat -r "desc(symlink)" b/a
  b/a: no such file in rev bc151a1f53bd
  [1]

#endif


unbundle tampered bundle

  $ hg init target
  $ cd target
  $ hg unbundle "$TESTDIR/bundles/tampered.hg"
  adding changesets
  adding manifests
  adding file changes
  added 5 changesets with 6 changes to 6 files (+4 heads)
  new changesets b7da9bf6b037:fc1393d727bc (5 drafts)
  (run 'hg heads' to see heads, 'hg merge' to merge)

attack .hg/test

  $ hg manifest -r0
  .hg/test
  $ hg update -Cr0
  abort: path contains illegal component: .hg/test
  [255]

attack foo/.hg/test

  $ hg manifest -r1
  foo/.hg/test
  $ hg update -Cr1
  abort: path 'foo/.hg/test' is inside nested repo 'foo'
  [255]

attack back/test where back symlinks to ..

  $ hg manifest -r2
  back
  back/test
#if symlink
  $ hg update -Cr2
  abort: path 'back/test' traverses symbolic link 'back'
  [255]
#else
('back' will be a file and cause some other system specific error)
  $ hg update -Cr2
  abort: $TESTTMP/target/back/test: $ENOTDIR$
  [255]
#endif

attack ../test

  $ hg manifest -r3
  ../test
  $ mkdir ../test
  $ echo data > ../test/file
  $ hg update -Cr3
  abort: path contains illegal component: ../test
  [255]
  $ cat ../test/file
  data

attack /tmp/test

  $ hg manifest -r4
  /tmp/test
  $ hg update -Cr4
  abort: path contains illegal component: /tmp/test
  [255]

  $ cd ..

Test symlink traversal on merge:
--------------------------------

#if symlink

set up symlink hell

  $ mkdir merge-symlink-out
  $ hg init merge-symlink
  $ cd merge-symlink
  $ touch base
  $ hg commit -qAm base
  $ ln -s ../merge-symlink-out a
  $ hg commit -qAm 'symlink a -> ../merge-symlink-out'
  $ hg up -q 0
  $ mkdir a
  $ touch a/poisoned
  $ hg commit -qAm 'file a/poisoned'
  $ hg log -G -T '{rev}: {desc}\n'
  @  2: file a/poisoned
  |
  | o  1: symlink a -> ../merge-symlink-out
  |/
  o  0: base
  

try trivial merge

  $ hg up -qC 1
  $ hg merge 2
  abort: path 'a/poisoned' traverses symbolic link 'a'
  [255]

try rebase onto other revision: cache of audited paths should be discarded,
and the rebase should fail (issue5628)

  $ hg up -qC 2
  $ hg rebase -s 2 -d 1 --config extensions.rebase=
  rebasing 2:e73c21d6b244 "file a/poisoned" (tip)
  abort: path 'a/poisoned' traverses symbolic link 'a'
  [255]
  $ ls ../merge-symlink-out

  $ cd ..

Test symlink traversal on update:
---------------------------------

  $ mkdir update-symlink-out
  $ hg init update-symlink
  $ cd update-symlink
  $ ln -s ../update-symlink-out a
  $ hg commit -qAm 'symlink a -> ../update-symlink-out'
  $ hg rm a
  $ mkdir a && touch a/b
  $ hg ci -qAm 'file a/b' a/b
  $ hg up -qC 0
  $ hg rm a
  $ mkdir a && touch a/c
  $ hg ci -qAm 'rm a, file a/c'
  $ hg log -G -T '{rev}: {desc}\n'
  @  2: rm a, file a/c
  |
  | o  1: file a/b
  |/
  o  0: symlink a -> ../update-symlink-out
  

try linear update where symlink already exists:

  $ hg up -qC 0
  $ hg up 1
  abort: path 'a/b' traverses symbolic link 'a'
  [255]

try linear update including symlinked directory and its content: paths are
audited first by calculateupdates(), where no symlink is created so both
'a' and 'a/b' are taken as good paths. still applyupdates() should fail.

  $ hg up -qC null
  $ hg up 1
  abort: path 'a/b' traverses symbolic link 'a'
  [255]
  $ ls ../update-symlink-out

try branch update replacing directory with symlink, and its content: the
path 'a' is audited as a directory first, which should be audited again as
a symlink.

  $ rm -f a
  $ hg up -qC 2
  $ hg up 1
  abort: path 'a/b' traverses symbolic link 'a'
  [255]
  $ ls ../update-symlink-out

  $ cd ..

#endif