annotate tests/test-verify-repo-operations.py @ 51681:522b4d729e89

mmap: populate the mapping by default

Without pre-population, accessing all data through a mmap can result in many
pagefault, reducing performance significantly. If the mmap is prepopulated, the
performance can no longer get slower than a full read.
(See benchmark number below)

In some cases were very few data is read, prepopulating can be overkill and
slower than populating on access (through page fault). So that behavior can be
controlled when the caller can pre-determine the best behavior.
(See benchmark number below)

In addition, testing with populating in a secondary thread yield great result
combining the best of each approach. This might be implemented in later
changesets.

In all cases, using mmap has a great effect on memory usage when many processes
run in parallel on the same machine.

### Benchmarks

# What did I run

A couple of month back I ran a large benchmark campaign to assess the impact of
various approach for using mmap with the revlog (and other files), it
highlighted a few benchmarks that capture the impact of the changes well. So to
validate this change I checked the following:

- log command displaying various revisions
  (read the changelog index)
- log command displaying the patch of listed revisions
  (read the changelog index, the manifest index and a few files indexes)
- unbundling a few revisions
  (read and write changelog, manifest and few files indexes, and walk the graph
  to update some cache)
- pushing a few revisions
  (read and write changelog, manifest and few files indexes, walk the graph to
  update some cache, performs various accesses locally and remotely during
  discovery)

Benchmarks were run using the default module policy (c+py) and the rust one. No
significant difference were found between the two implementation, so we will
present result using the default policy (unless otherwise specified).

I ran them on a few repositories :

- mercurial: a "public changeset only" copy of mercurial from 2018-08-01 using
  zstd compression and sparse-revlog

- pypy: a copy of pypy from 2018-08-01 using zstd compression and sparse-revlog

- netbeans: a copy of netbeans from 2018-08-01 using zstd compression and
  sparse-revlog

- mozilla-try: a copy of mozilla-try from 2019-02-18 using zstd compression and
  sparse-revlog

- mozilla-try persistent-nodemap: Same as the above but with a persistent
  nodemap. Used for the log --patch benchmark only

# Results

For the smaller repositories (mercurial, pypy), the impact of mmap is almost
imperceptible, other cost dominating the operation. The impact of prepopulating
is undiscernible in the benchmark we ran.

For larger repositories the benchmark support explanation given above:

On netbeans, the log can be about 1% faster without repopulation (for a
difference < 100ms) but unbundle becomes a bit slower, even when small.

### data-env-vars.name                             = netbeans-2018-08-01-zstd-sparse-revlog
  # benchmark.name                                 = hg.command.unbundle
  # benchmark.variants.issue6528                   = disabled
  # benchmark.variants.reuse-external-delta-parent = yes
  # benchmark.variants.revs                        = any-1-extra-rev
  # benchmark.variants.source                      = unbundle
  # benchmark.variants.verbosity                   = quiet
with-populate: 0.240157
no-populate:   0.265087  (+10.38%, +0.02)
  # benchmark.variants.revs                        = any-100-extra-rev
with-populate: 1.459518
no-populate:   1.481290   (+1.49%, +0.02)
 ## benchmark.name                                 = hg.command.push
  # benchmark.variants.explicit-rev                = none
  # benchmark.variants.issue6528                   = disabled
  # benchmark.variants.protocol                    = ssh
  # benchmark.variants.reuse-external-delta-parent = yes
  # benchmark.variants.revs                        = any-1-extra-rev
with-populate: 0.771919
no-populate:   0.792025   (+2.60%, +0.02)
  # benchmark.variants.revs                        = any-100-extra-rev
with-populate: 1.459518
no-populate:   1.481290   (+1.49%, +0.02)


For mozilla-try, the "slow down" from pre-populate for small `hg log` is more
visible, but still small in absolute time. (using rust value for the persistent
nodemap value to be relevant).

### data-env-vars.name            = mozilla-try-2019-02-18-ds2-pnm
  # benchmark.name                = hg.command.log
  # bin-env-vars.hg.flavor        = rust
  # benchmark.variants.patch      = yes
  # benchmark.variants.limit-rev  = 1
with-populate: 0.237813
no-populate:   0.229452   (-3.52%, -0.01)
  # benchmark.variants.limit-rev  = 10
  # benchmark.variants.patch      = yes
with-populate: 1.213578
no-populate:   1.205189
### data-env-vars.name            = mozilla-try-2019-02-18-zstd-sparse-revlog
  # benchmark.variants.limit-rev  = 1000
  # benchmark.variants.patch      = no
  # benchmark.variants.rev        = tip
with-populate: 0.198607
no-populate:   0.195038   (-1.80%, -0.00)

However pre-populating provide a significant boost on more complex operations
like unbundle or push:

### data-env-vars.name                             = mozilla-try-2019-02-18-zstd-sparse-revlog
  # benchmark.name                                 = hg.command.push
  # benchmark.variants.explicit-rev                = none
  # benchmark.variants.issue6528                   = disabled
  # benchmark.variants.protocol                    = ssh
  # benchmark.variants.reuse-external-delta-parent = yes
  # benchmark.variants.revs                        = any-1-extra-rev
with-populate: 4.798632
no-populate:   4.953295   (+3.22%, +0.15)
  # benchmark.variants.revs                        = any-100-extra-rev
with-populate: 4.903618
no-populate:   5.014963   (+2.27%, +0.11)
 ## benchmark.name                                 = hg.command.unbundle
  # benchmark.variants.revs                        = any-1-extra-rev
with-populate: 1.423411
no-populate:   1.585365  (+11.38%, +0.16)
  # benchmark.variants.revs                        = any-100-extra-rev
with-populate: 1.537909
no-populate:   1.688489   (+9.79%, +0.15)

"""Fuzz testing for operations against a Mercurial repository

This uses Hypothesis's stateful testing to generate random repository
operations and test Mercurial using them, both to see if there are any
unexpected errors and to compare different versions of it."""

import os
import subprocess
import sys

# Only run if slow tests are allowed
if subprocess.call(
    [os.environ['PYTHON'], '%s/hghave' % os.environ['TESTDIR'], 'slow']
):
    sys.exit(80)

# These tests require Hypothesis and pytz to be installed.
# Running 'pip install hypothesis pytz' will achieve that.
# Note: This won't work if you're running Python < 2.7.
try:
    from hypothesis.extra.datetime import datetimes
except ImportError:
    sys.stderr.write("skipped: hypothesis or pytz not installed" + os.linesep)
    sys.exit(80)

# If you are running an old version of pip you may find that the enum34
# backport is not installed automatically. If so 'pip install enum34' will
# fix this problem.
try:
    import enum

    assert enum  # Silence pyflakes
except ImportError:
    sys.stderr.write("skipped: enum34 not installed" + os.linesep)
    sys.exit(80)

import binascii
from contextlib import contextmanager
import pipes
import shutil
import silenttestrunner
import subprocess

from hypothesis.errors import HypothesisException
from hypothesis.stateful import (
    rule,
    RuleBasedStateMachine,
    Bundle,
    precondition,
)
from hypothesis import settings, note, strategies as st
from hypothesis.configuration import set_hypothesis_home_dir
from hypothesis.database import ExampleDatabase

testdir = os.path.abspath(os.environ["TESTDIR"])

# We store Hypothesis examples here rather in the temporary test directory
# so that when rerunning a failing test this always results in refinding the
# previous failure. This directory is in .hgignore and should not be checked in
# but is useful to have for development.
set_hypothesis_home_dir(os.path.join(testdir, ".hypothesis"))

runtests = os.path.join(os.environ["RUNTESTDIR"], "run-tests.py")
testtmp = os.environ["TESTTMP"]
assert os.path.isdir(testtmp)

generatedtests = os.path.join(testdir, "hypothesis-generated")

try:
    os.makedirs(generatedtests)
except OSError:
    pass

# We write out generated .t files to a file in order to ease debugging and to
# give a starting point for turning failures Hypothesis finds into normal
# tests. In order to ensure that multiple copies of this test can be run in
# parallel we use atomic file create to ensure that we always get a unique
# name.
file_index = 0
while True:
    file_index += 1
    savefile = os.path.join(
        generatedtests, "test-generated-%d.t" % (file_index,)
    )
    try:
        os.close(os.open(savefile, os.O_CREAT | os.O_EXCL | os.O_WRONLY))
        break
    except FileExistsError:
        pass
assert os.path.exists(savefile)

hgrc = os.path.join(".hg", "hgrc")

filecharacters = (
    "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"
    "[]^_`;=@{}~ !#$%&'()+,-"
)

files = (
    st.text(filecharacters, min_size=1)
    .map(lambda x: x.strip())
    .filter(bool)
    .map(lambda s: s.encode('ascii'))
)

safetext = st.text(
    st.characters(
        min_codepoint=1, max_codepoint=127, blacklist_categories=('Cc', 'Cs')
    ),
    min_size=1,
).map(lambda s: s.encode('utf-8'))

extensions = st.sampled_from(
    (
        'shelve',
        'mq',
        'blackbox',
    )
)


@contextmanager
def acceptableerrors(*args):
    """Sometimes we know an operation we're about to perform might fail, and
    we're OK with some of the failures. In those cases this may be used as a
    context manager and will swallow expected failures, as identified by
    substrings of the error message Mercurial emits."""
    try:
        yield
    except subprocess.CalledProcessError as e:
        if not any(a in e.output for a in args):
            note(e.output)
            raise


reponames = st.text("abcdefghijklmnopqrstuvwxyz01234556789", min_size=1).map(
    lambda s: s.encode('ascii')
)


class verifyingstatemachine(RuleBasedStateMachine):
    """This defines the set of acceptable operations on a Mercurial repository
    using Hypothesis's RuleBasedStateMachine.

    The general concept is that we manage multiple repositories inside a
    repos/ directory in our temporary test location. Some of these are freshly
    inited, some are clones of the others. Our current working directory is
    always inside one of these repositories while the tests are running.

    Hypothesis then performs a series of operations against these repositories,
    including hg commands, generating contents and editing the .hgrc file.
    If these operations fail in unexpected ways or behave differently in
    different configurations of Mercurial, the test will fail and a minimized
    .t test file will be written to the hypothesis-generated directory to
    exhibit that failure.

    Operations are defined as methods with @rule() decorators. See the
    Hypothesis documentation at
    http://hypothesis.readthedocs.org/en/release/stateful.html for more
    details."""

    # A bundle is a reusable collection of previously generated data which may
    # be provided as arguments to future operations.
    repos = Bundle('repos')
    paths = Bundle('paths')
    contents = Bundle('contents')
    branches = Bundle('branches')
    committimes = Bundle('committimes')

    def __init__(self):
        super(verifyingstatemachine, self).__init__()
        self.repodir = os.path.join(testtmp, "repos")
        if os.path.exists(self.repodir):
            shutil.rmtree(self.repodir)
        os.chdir(testtmp)
        self.log = []
        self.failed = False
        self.configperrepo = {}
        self.all_extensions = set()
        self.non_skippable_extensions = set()

        self.mkdirp("repos")
        self.cd("repos")
        self.mkdirp("repo1")
        self.cd("repo1")
        self.hg("init")

    def teardown(self):
        """On teardown we clean up after ourselves as usual, but we also
        do some additional testing: We generate a .t file based on our test
        run using run-test.py -i to get the correct output.

        We then test it in a number of other configurations, verifying that
        each passes the same test."""
        super(verifyingstatemachine, self).teardown()
        try:
            shutil.rmtree(self.repodir)
        except OSError:
            pass
        ttest = os.linesep.join("  " + l for l in self.log)
        os.chdir(testtmp)
        path = os.path.join(testtmp, "test-generated.t")
        with open(path, 'w') as o:
            o.write(ttest + os.linesep)
        with open(os.devnull, "w") as devnull:
            rewriter = subprocess.Popen(
                [runtests, "--local", "-i", path],
                stdin=subprocess.PIPE,
                stdout=devnull,
                stderr=devnull,
            )
            rewriter.communicate("yes")
            with open(path, 'r') as i:
                ttest = i.read()

        e = None
        if not self.failed:
            try:
                output = subprocess.check_output(
                    [runtests, path, "--local", "--pure"],
                    stderr=subprocess.STDOUT,
                )
                assert "Ran 1 test" in output, output
                for ext in self.all_extensions - self.non_skippable_extensions:
                    tf = os.path.join(
                        testtmp, "test-generated-no-%s.t" % (ext,)
                    )
                    with open(tf, 'w') as o:
                        for l in ttest.splitlines():
                            if l.startswith("  $ hg"):
                                l = l.replace(
                                    "--config %s=" % (extensionconfigkey(ext),),
                                    "",
                                )
                            o.write(l + os.linesep)
                    with open(tf, 'r') as r:
                        t = r.read()
                        assert ext not in t, t
                    output = subprocess.check_output(
                        [
                            runtests,
                            tf,
                            "--local",
                        ],
                        stderr=subprocess.STDOUT,
                    )
                    assert "Ran 1 test" in output, output
            except subprocess.CalledProcessError as e:
                note(e.output)
        if self.failed or e is not None:
            with open(savefile, "wb") as o:
                o.write(ttest)
        if e is not None:
            raise e

    def execute_step(self, step):
        try:
            return super(verifyingstatemachine, self).execute_step(step)
        except (HypothesisException, KeyboardInterrupt):
            raise
        except Exception:
            self.failed = True
            raise

    # Section: Basic commands.
    def mkdirp(self, path):
        if os.path.exists(path):
            return
        self.log.append(
            "$ mkdir -p -- %s" % (pipes.quote(os.path.relpath(path)),)
        )
        os.makedirs(path)

    def cd(self, path):
        path = os.path.relpath(path)
        if path == ".":
            return
        os.chdir(path)
        self.log.append("$ cd -- %s" % (pipes.quote(path),))

    def hg(self, *args):
        extra_flags = []
        for key, value in self.config.items():
            extra_flags.append("--config")
            extra_flags.append("%s=%s" % (key, value))
        self.command("hg", *(tuple(extra_flags) + args))

    def command(self, *args):
        self.log.append("$ " + ' '.join(map(pipes.quote, args)))
        subprocess.check_output(args, stderr=subprocess.STDOUT)

    # Section: Set up basic data
    # This section has no side effects but generates data that we will want
    # to use later.
    @rule(
        target=paths,
        source=st.lists(files, min_size=1).map(lambda l: os.path.join(*l)),
    )
    def genpath(self, source):
        return source

    @rule(
        target=committimes,
        when=datetimes(min_year=1970, max_year=2038) | st.none(),
    )
    def gentime(self, when):
        return when

    @rule(
        target=contents,
        content=st.one_of(
            st.binary(), st.text().map(lambda x: x.encode('utf-8'))
        ),
    )
    def gencontent(self, content):
        return content

    @rule(
        target=branches,
        name=safetext,
    )
    def genbranch(self, name):
        return name

    @rule(target=paths, source=paths)
    def lowerpath(self, source):
        return source.lower()

    @rule(target=paths, source=paths)
    def upperpath(self, source):
        return source.upper()

    # Section: Basic path operations
    @rule(path=paths, content=contents)
    def writecontent(self, path, content):
        self.unadded_changes = True
        if os.path.isdir(path):
            return
        parent = os.path.dirname(path)
        if parent:
            try:
                self.mkdirp(parent)
            except OSError:
                # It may be the case that there is a regular file that has
                # previously been created that has the same name as an ancestor
                # of the current path. This will cause mkdirp to fail with this
                # error. We just turn this into a no-op in that case.
                return
        with open(path, 'wb') as o:
            o.write(content)
        self.log.append(
            (
                "$ \"$PYTHON\" -c 'import binascii; "
                "print(binascii.unhexlify(\"%s\"))' > %s"
            )
            % (
                binascii.hexlify(content),
                pipes.quote(path),
            )
        )

    @rule(path=paths)
    def addpath(self, path):
        if os.path.exists(path):
            self.hg("add", "--", path)

    @rule(path=paths)
    def forgetpath(self, path):
        if os.path.exists(path):
            with acceptableerrors(
                "file is already untracked",
            ):
                self.hg("forget", "--", path)

    @rule(s=st.none() | st.integers(0, 100))
    def addremove(self, s):
        args = ["addremove"]
        if s is not None:
            args.extend(["-s", str(s)])
        self.hg(*args)

    @rule(path=paths)
    def removepath(self, path):
        if os.path.exists(path):
            with acceptableerrors(
                'file is untracked',
                'file has been marked for add',
                'file is modified',
            ):
                self.hg("remove", "--", path)

    @rule(
        message=safetext,
        amend=st.booleans(),
        when=committimes,
        addremove=st.booleans(),
        secret=st.booleans(),
        close_branch=st.booleans(),
    )
    def maybecommit(
        self, message, amend, when, addremove, secret, close_branch
    ):
        command = ["commit"]
        errors = ["nothing changed"]
        if amend:
            errors.append("cannot amend public changesets")
            command.append("--amend")
        command.append("-m" + pipes.quote(message))
        if secret:
            command.append("--secret")
        if close_branch:
            command.append("--close-branch")
            errors.append("can only close branch heads")
        if addremove:
            command.append("--addremove")
        if when is not None:
            if when.year == 1970:
                errors.append('negative date value')
            if when.year == 2038:
                errors.append('exceeds 32 bits')
            command.append(
                "--date=%s" % (when.strftime('%Y-%m-%d %H:%M:%S %z'),)
            )

        with acceptableerrors(*errors):
            self.hg(*command)

    # Section: Repository management
    @property
    def currentrepo(self):
        return os.path.basename(os.getcwd())

    @property
    def config(self):
        return self.configperrepo.setdefault(self.currentrepo, {})

    @rule(
        target=repos,
        source=repos,
        name=reponames,
    )
    def clone(self, source, name):
        if not os.path.exists(os.path.join("..", name)):
            self.cd("..")
            self.hg("clone", source, name)
            self.cd(name)
        return name

    @rule(
        target=repos,
        name=reponames,
    )
    def fresh(self, name):
        if not os.path.exists(os.path.join("..", name)):
            self.cd("..")
            self.mkdirp(name)
            self.cd(name)
            self.hg("init")
        return name

    @rule(name=repos)
    def switch(self, name):
        self.cd(os.path.join("..", name))
        assert self.currentrepo == name
        assert os.path.exists(".hg")

    @rule(target=repos)
    def origin(self):
        return "repo1"

    @rule()
    def pull(self, repo=repos):
        with acceptableerrors(
            "repository default not found",
            "repository is unrelated",
        ):
            self.hg("pull")

    @rule(newbranch=st.booleans())
    def push(self, newbranch):
        with acceptableerrors(
            "default repository not configured",
            "no changes found",
        ):
            if newbranch:
                self.hg("push", "--new-branch")
            else:
                with acceptableerrors("creates new branches"):
                    self.hg("push")

    # Section: Simple side effect free "check" operations
    @rule()
    def log(self):
        self.hg("log")

    @rule()
    def verify(self):
        self.hg("verify")

    @rule()
    def diff(self):
        self.hg("diff", "--nodates")

    @rule()
    def status(self):
        self.hg("status")

    @rule()
    def export(self):
        self.hg("export")

    # Section: Branch management
    @rule()
    def checkbranch(self):
        self.hg("branch")

    @rule(branch=branches)
    def switchbranch(self, branch):
        with acceptableerrors(
            'cannot use an integer as a name',
            'cannot be used in a name',
            'a branch of the same name already exists',
            'is reserved',
        ):
            self.hg("branch", "--", branch)

    @rule(branch=branches, clean=st.booleans())
    def update(self, branch, clean):
        with acceptableerrors(
            'unknown revision',
            'parse error',
        ):
            if clean:
                self.hg("update", "-C", "--", branch)
            else:
                self.hg("update", "--", branch)

    # Section: Extension management
    def hasextension(self, extension):
        return extensionconfigkey(extension) in self.config

    def commandused(self, extension):
        assert extension in self.all_extensions
        self.non_skippable_extensions.add(extension)

    @rule(extension=extensions)
    def addextension(self, extension):
        self.all_extensions.add(extension)
        self.config[extensionconfigkey(extension)] = ""

    @rule(extension=extensions)
    def removeextension(self, extension):
        self.config.pop(extensionconfigkey(extension), None)

    # Section: Commands from the shelve extension
    @rule()
    @precondition(lambda self: self.hasextension("shelve"))
    def shelve(self):
        self.commandused("shelve")
        with acceptableerrors("nothing changed"):
            self.hg("shelve")

    @rule()
    @precondition(lambda self: self.hasextension("shelve"))
    def unshelve(self):
        self.commandused("shelve")
        with acceptableerrors("no shelved changes to apply"):
            self.hg("unshelve")


class writeonlydatabase(ExampleDatabase):
    def __init__(self, underlying):
        super(ExampleDatabase, self).__init__()
        self.underlying = underlying

    def fetch(self, key):
        return ()

    def save(self, key, value):
        self.underlying.save(key, value)

    def delete(self, key, value):
        self.underlying.delete(key, value)

    def close(self):
        self.underlying.close()


def extensionconfigkey(extension):
    return "extensions." + extension


settings.register_profile(
    'default',
    settings(
        timeout=300,
        stateful_step_count=50,
        max_examples=10,
    ),
)

settings.register_profile(
    'fast',
    settings(
        timeout=10,
        stateful_step_count=20,
        max_examples=5,
        min_satisfying_examples=1,
        max_shrinks=0,
    ),
)

settings.register_profile(
    'continuous',
    settings(
        timeout=-1,
        stateful_step_count=1000,
        max_examples=10 ** 8,
        max_iterations=10 ** 8,
        database=writeonlydatabase(settings.default.database),
    ),
)

settings.load_profile(os.getenv('HYPOTHESIS_PROFILE', 'default'))

verifyingtest = verifyingstatemachine.TestCase

verifyingtest.settings = settings.default

if __name__ == '__main__':
    try:
        silenttestrunner.main(__name__)
    finally:
        # So as to prevent proliferation of useless test files, if we never
        # actually wrote a failing test we clean up after ourselves and delete
        # the file for doing so that we owned.
        if os.path.exists(savefile) and os.path.getsize(savefile) == 0:
            os.unlink(savefile)