view mercurial/repocache.py @ 48687:f8f2ecdde4b5

branchmap: skip obsolete revisions while computing heads It's time to make this part of core Mercurial obsolescence-aware. Not considering obsolete revisions when computing heads is clearly what Mercurial should do. But there are a couple of small issues: - Let's say tip of the repo is obsolete. There are two ways of finding tiprev for branchcache (both are in use): looking at input data for update() and looking at computed heads after update(). Previously, repo tip would be tiprev of the branchcache. With this patch, an obsolete revision can no longer be tiprev. And depending on what way we use for finding tiprev (input data vs computed heads) we'll get a different result. This is relevant when recomputing cache key from cache contents, and may lead to updating cache for obsolete revisions multiple times (not from scratch, because it still would be considered valid for a subset of revisions in the repo). - If all commits on a branch are obsolete, the branchcache will include that branch, but the list of heads will be empty (that's why there's now `if not heads` when recomputing tiprev/tipnode from cache contents). Having an entry for every branch is currently required for notify extension (and test-notify.t to pass), because notify doesn't handle revsets in its subscription config very well and will throw an error if e.g. a branch doesn't exist. - Cloning static HTTP repos may try to stat() a non-existent obsstore file. The issue is that we now care about obsolescence during clone, but statichttpvfs doesn't implement a stat method, so a regular vfs.stat() is used, and it assumes that file is local and calls os.stat(). During a clone, we're trying to stat() .hg/store/obsstore, but in static HTTP case we provide a literal URL to the obsstore file on the remote as if it were a local file path. On windows it actually results in a failure in test-static-http.t. The first issue is going to be addressed in a series dedicated to making sure branchcache is properly and timely written on disk (it wasn't perfect even before this patch, but there aren't enough tests to demonstrate that). The second issue will be addressed in a future patch for notify extension that will make it not raise an exception if a branch doesn't exist. And the third one was partially addressed in the previous patch in this series and will be properly fixed in a future patch when this series is accepted. filteredhash() grows a keyword argument to make sure that branchcache is also invalidated when there are new obsolete revisions in its repo view. This way the on-disk cache format is unchanged and compatible between versions (although it will obviously be recomputed when switching versions before/after this patch and the repo has obsolete revisions). There's one test that uses plain `hg up` without arguments while updated to a pruned commit. To make this test pass, simply return current working directory parent. Later in this series this code will be replaced by what prune command does: updating to the closest non-obsolete ancestor. Test changes: test-branch-change.t: update branch head and cache update message. The head of default listed in hg heads is changed because revision 2 was rewritten as 7, and 1 is the closest ancestor on the same branch, so it's the head of default now. The cache invalidation message appears now because of the cache hash change, since we're now accounting for obsolete revisions. Here's some context: "served.hidden" repo filter means everything is visible (no filtered revisions), so before this series branch2-served.hidden file would not contain any cache hash, only revnum and node. Now it also has a hash when there are obsolete changesets in the repo. The command that the message appears for is changing branch of 5 and 6, which are now obsolete, so the cache hash changes. In general, when cache is simply out-of-date, it can be updated using the old version as a base. But if cache hash differs, then the cache for that particular repo filter is recomputed (at least with the current implementation). This is what happens here. test-obsmarker-template.t: the pull reports 2 heads changed, but after that the repo correctly sees only 1. The new message could be better, but it's still an improvement over the previous one where hg pull suggested merging with an obsolete revision. test-obsolete.t: we can see these revisions in hg log --hidden, but they shouldn't be considered heads even with --hidden. test-rebase-obsolete{,2}.t: there were new heads created previously after making new orphan changesets, but they weren't detected. Now we are properly detecting and reporting them. test-rebase-obsolete4.t: there's only one head now because the other head is pruned and was falsely reported before. test-static-http.t: add obsstore to the list of requested files. This file doesn't exist on the remotes, but clients want it anyway (they get 404). This is fine, because there are other nonexistent files that clients request, like .hg/bookmarks or .hg/cache/tags2-served. Differential Revision: https://phab.mercurial-scm.org/D12097
author Anton Shestakov <av6@dwimlabs.net>
date Fri, 07 Jan 2022 11:53:23 +0300
parents 8ff1ecfadcd1
children 6000f5b25c9b
line wrap: on
line source

# repocache.py - in-memory repository cache for long-running services
#
# Copyright 2018 Yuya Nishihara <yuya@tcha.org>
#
# This software may be used and distributed according to the terms of the
# GNU General Public License version 2 or any later version.

from __future__ import absolute_import

import collections
import gc
import threading

from . import (
    error,
    hg,
    obsolete,
    scmutil,
    util,
)


class repoloader(object):
    """Load repositories in background thread

    This is designed for a forking server. A cached repo cannot be obtained
    until the server fork()s a worker and the loader thread stops.
    """

    def __init__(self, ui, maxlen):
        self._ui = ui.copy()
        self._cache = util.lrucachedict(max=maxlen)
        # use deque and Event instead of Queue since deque can discard
        # old items to keep at most maxlen items.
        self._inqueue = collections.deque(maxlen=maxlen)
        self._accepting = False
        self._newentry = threading.Event()
        self._thread = None

    def start(self):
        assert not self._thread
        if self._inqueue.maxlen == 0:
            # no need to spawn loader thread as the cache is disabled
            return
        self._accepting = True
        self._thread = threading.Thread(target=self._mainloop)
        self._thread.start()

    def stop(self):
        if not self._thread:
            return
        self._accepting = False
        self._newentry.set()
        self._thread.join()
        self._thread = None
        self._cache.clear()
        self._inqueue.clear()

    def load(self, path):
        """Request to load the specified repository in background"""
        self._inqueue.append(path)
        self._newentry.set()

    def get(self, path):
        """Return a cached repo if available

        This function must be called after fork(), where the loader thread
        is stopped. Otherwise, the returned repo might be updated by the
        loader thread.
        """
        if self._thread and self._thread.is_alive():
            raise error.ProgrammingError(
                b'cannot obtain cached repo while loader is active'
            )
        return self._cache.peek(path, None)

    def _mainloop(self):
        while self._accepting:
            # Avoid heavy GC after fork(), which would cancel the benefit of
            # COW. We assume that GIL is acquired while GC is underway in the
            # loader thread. If that isn't true, we might have to move
            # gc.collect() to the main thread so that fork() would never stop
            # the thread where GC is in progress.
            gc.collect()

            self._newentry.wait()
            while self._accepting:
                self._newentry.clear()
                try:
                    path = self._inqueue.popleft()
                except IndexError:
                    break
                scmutil.callcatch(self._ui, lambda: self._load(path))

    def _load(self, path):
        start = util.timer()
        # TODO: repo should be recreated if storage configuration changed
        try:
            # pop before loading so inconsistent state wouldn't be exposed
            repo = self._cache.pop(path)
        except KeyError:
            repo = hg.repository(self._ui, path).unfiltered()
        _warmupcache(repo)
        repo.ui.log(
            b'repocache',
            b'loaded repo into cache: %s (in %.3fs)\n',
            path,
            util.timer() - start,
        )
        self._cache.insert(path, repo)


# TODO: think about proper API of preloading cache
def _warmupcache(repo):
    repo.invalidateall()
    repo.changelog
    repo.obsstore._all
    repo.obsstore.successors
    repo.obsstore.predecessors
    repo.obsstore.children
    for name in obsolete.cachefuncs:
        obsolete.getrevs(repo, name)
    repo._phasecache.loadphaserevs(repo)


# TODO: think about proper API of attaching preloaded attributes
def copycache(srcrepo, destrepo):
    """Copy cached attributes from srcrepo to destrepo"""
    destfilecache = destrepo._filecache
    srcfilecache = srcrepo._filecache
    if b'changelog' in srcfilecache:
        destfilecache[b'changelog'] = ce = srcfilecache[b'changelog']
        ce.obj.opener = ce.obj._realopener = destrepo.svfs
    if b'obsstore' in srcfilecache:
        destfilecache[b'obsstore'] = ce = srcfilecache[b'obsstore']
        ce.obj.svfs = destrepo.svfs
    if b'_phasecache' in srcfilecache:
        destfilecache[b'_phasecache'] = ce = srcfilecache[b'_phasecache']
        ce.obj.opener = destrepo.svfs