mercurial/hgweb/hgwebdir_mod.py

hgwebdir: avoid systematic full garbage collection

Forcing a systematic full garbage collection upon each request
can serioulsy harm performance. This is reported as
https://bz.mercurial-scm.org/show_bug.cgi?id=6075

With this change we're performing the full collection according
to a new setting, `experimental.web.full-garbage-collection-rate`.
The default value is 1, which doesn't change the behavior and will
allow us to test on real use cases. If the value is 0, no full garbage
collection occurs.

Regardless of the value of the setting, a partial garbage collection
still occurs upon each request (not attempting to collect objects from
the oldest generation). This should be enough to take care of
reference cycles that have been created by the last request
(assessment of this requires changing the setting, not to be 1).

In my experience chasing memory leaks in Mercurial servers,
the full collection never reclaimed any memory, but this is with
Python 3 and biased towards small repositories.

On the other hand, as explained in the Python developer docs [1],
frequent full collections are very harmful in terms of performance if
lots of objects survive the collection, and hence stay in the
oldest generation. Note that `gc.collect()` is indeed trying to
collect the oldest generation [2]. This happens usually in two cases:

- unwanted lingering objects (i.e., an actual memory leak that
  the GC cannot do anything about). Sadly, we have lots of those
  these days.
- desireable long-term objects, typically in caches (not inner caches
  carried by repositories, which should be collected with them). This
  is a subject of interest for the Heptapod project.

In short, the flat rate that this change still permits is
probably a bad idea in most cases, and the default value can
be tweaked later on (or even be set to 0) according to experiments
in the wild.

The test is inspired from test-hgwebdir-paths.py

[1] https://devguide.python.org/garbage_collector/#collecting-the-oldest-generation
[2] https://docs.python.org/3/library/gc.html#gc.collect

Differential Revision: https://phab.mercurial-scm.org/D11204

# hgweb/hgwebdir_mod.py - Web interface for a directory of repositories.
#
# Copyright 21 May 2005 - (c) 2005 Jake Edge <[e-mail]>
# Copyright 2005, 2006 Olivia Mackall <[e-mail]>
#
# 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 gc
import os
import time

from ..i18n import _

from .common import (
    ErrorResponse,
    HTTP_SERVER_ERROR,
    cspvalues,
    get_contact,
    get_mtime,
    ismember,
    paritygen,
    staticfile,
    statusmessage,
)

from .. import (
    configitems,
    encoding,
    error,
    extensions,
    hg,
    pathutil,
    profiling,
    pycompat,
    rcutil,
    registrar,
    scmutil,
    templater,
    templateutil,
    ui as uimod,
    util,
)

from . import (
    hgweb_mod,
    request as requestmod,
    webutil,
    wsgicgi,
)
from ..utils import dateutil


def cleannames(items):
    return [(util.pconvert(name).strip(b'/'), path) for name, path in items]


def findrepos(paths):
    repos = []
    for prefix, root in cleannames(paths):
        roothead, roottail = os.path.split(root)
        # "foo = /bar/*" or "foo = /bar/**" lets every repo /bar/N in or below
        # /bar/ be served as as foo/N .
        # '*' will not search inside dirs with .hg (except .hg/patches),
        # '**' will search inside dirs with .hg (and thus also find subrepos).
        try:
            recurse = {b'*': False, b'**': True}[roottail]
        except KeyError:
            repos.append((prefix, root))
            continue
        roothead = os.path.normpath(util.abspath(roothead))
        paths = scmutil.walkrepos(roothead, followsym=True, recurse=recurse)
        repos.extend(urlrepos(prefix, roothead, paths))
    return repos


def urlrepos(prefix, roothead, paths):
    """yield url paths and filesystem paths from a list of repo paths

    >>> conv = lambda seq: [(v, util.pconvert(p)) for v,p in seq]
    >>> conv(urlrepos(b'hg', b'/opt', [b'/opt/r', b'/opt/r/r', b'/opt']))
    [('hg/r', '/opt/r'), ('hg/r/r', '/opt/r/r'), ('hg', '/opt')]
    >>> conv(urlrepos(b'', b'/opt', [b'/opt/r', b'/opt/r/r', b'/opt']))
    [('r', '/opt/r'), ('r/r', '/opt/r/r'), ('', '/opt')]
    """
    for path in paths:
        path = os.path.normpath(path)
        yield (
            prefix + b'/' + util.pconvert(path[len(roothead) :]).lstrip(b'/')
        ).strip(b'/'), path


def readallowed(ui, req):
    """Check allow_read and deny_read config options of a repo's ui object
    to determine user permissions.  By default, with neither option set (or
    both empty), allow all users to read the repo.  There are two ways a
    user can be denied read access:  (1) deny_read is not empty, and the
    user is unauthenticated or deny_read contains user (or *), and (2)
    allow_read is not empty and the user is not in allow_read.  Return True
    if user is allowed to read the repo, else return False."""

    user = req.remoteuser

    deny_read = ui.configlist(b'web', b'deny_read', untrusted=True)
    if deny_read and (not user or ismember(ui, user, deny_read)):
        return False

    allow_read = ui.configlist(b'web', b'allow_read', untrusted=True)
    # by default, allow reading if no allow_read option has been set
    if not allow_read or ismember(ui, user, allow_read):
        return True

    return False


def rawindexentries(ui, repos, req, subdir=b''):
    descend = ui.configbool(b'web', b'descend')
    collapse = ui.configbool(b'web', b'collapse')
    seenrepos = set()
    seendirs = set()
    for name, path in repos:

        if not name.startswith(subdir):
            continue
        name = name[len(subdir) :]
        directory = False

        if b'/' in name:
            if not descend:
                continue

            nameparts = name.split(b'/')
            rootname = nameparts[0]

            if not collapse:
                pass
            elif rootname in seendirs:
                continue
            elif rootname in seenrepos:
                pass
            else:
                directory = True
                name = rootname

                # redefine the path to refer to the directory
                discarded = b'/'.join(nameparts[1:])

                # remove name parts plus accompanying slash
                path = path[: -len(discarded) - 1]

                try:
                    hg.repository(ui, path)
                    directory = False
                except (IOError, error.RepoError):
                    pass

        parts = [
            req.apppath.strip(b'/'),
            subdir.strip(b'/'),
            name.strip(b'/'),
        ]
        url = b'/' + b'/'.join(p for p in parts if p) + b'/'

        # show either a directory entry or a repository
        if directory:
            # get the directory's time information
            try:
                d = (get_mtime(path), dateutil.makedate()[1])
            except OSError:
                continue

            # add '/' to the name to make it obvious that
            # the entry is a directory, not a regular repository
            row = {
                b'contact': b"",
                b'contact_sort': b"",
                b'name': name + b'/',
                b'name_sort': name,
                b'url': url,
                b'description': b"",
                b'description_sort': b"",
                b'lastchange': d,
                b'lastchange_sort': d[1] - d[0],
                b'archives': templateutil.mappinglist([]),
                b'isdirectory': True,
                b'labels': templateutil.hybridlist([], name=b'label'),
            }

            seendirs.add(name)
            yield row
            continue

        u = ui.copy()
        if rcutil.use_repo_hgrc():
            try:
                u.readconfig(os.path.join(path, b'.hg', b'hgrc'))
            except Exception as e:
                u.warn(_(b'error reading %s/.hg/hgrc: %s\n') % (path, e))
                continue

        def get(section, name, default=uimod._unset):
            return u.config(section, name, default, untrusted=True)

        if u.configbool(b"web", b"hidden", untrusted=True):
            continue

        if not readallowed(u, req):
            continue

        # update time with local timezone
        try:
            r = hg.repository(ui, path)
        except IOError:
            u.warn(_(b'error accessing repository at %s\n') % path)
            continue
        except error.RepoError:
            u.warn(_(b'error accessing repository at %s\n') % path)
            continue
        try:
            d = (get_mtime(r.spath), dateutil.makedate()[1])
        except OSError:
            continue

        contact = get_contact(get)
        description = get(b"web", b"description")
        seenrepos.add(name)
        name = get(b"web", b"name", name)
        labels = u.configlist(b'web', b'labels', untrusted=True)
        row = {
            b'contact': contact or b"unknown",
            b'contact_sort': contact.upper() or b"unknown",
            b'name': name,
            b'name_sort': name,
            b'url': url,
            b'description': description or b"unknown",
            b'description_sort': description.upper() or b"unknown",
            b'lastchange': d,
            b'lastchange_sort': d[1] - d[0],
            b'archives': webutil.archivelist(u, b"tip", url),
            b'isdirectory': None,
            b'labels': templateutil.hybridlist(labels, name=b'label'),
        }

        yield row


def _indexentriesgen(
    context, ui, repos, req, stripecount, sortcolumn, descending, subdir
):
    rows = rawindexentries(ui, repos, req, subdir=subdir)

    sortdefault = None, False

    if sortcolumn and sortdefault != (sortcolumn, descending):
        sortkey = b'%s_sort' % sortcolumn
        rows = sorted(rows, key=lambda x: x[sortkey], reverse=descending)

    for row, parity in zip(rows, paritygen(stripecount)):
        row[b'parity'] = parity
        yield row


def indexentries(
    ui, repos, req, stripecount, sortcolumn=b'', descending=False, subdir=b''
):
    args = (ui, repos, req, stripecount, sortcolumn, descending, subdir)
    return templateutil.mappinggenerator(_indexentriesgen, args=args)


class hgwebdir(object):
    """HTTP server for multiple repositories.

    Given a configuration, different repositories will be served depending
    on the request path.

    Instances are typically used as WSGI applications.
    """

    def __init__(self, conf, baseui=None):
        self.conf = conf
        self.baseui = baseui
        self.ui = None
        self.lastrefresh = 0
        self.motd = None
        self.refresh()
        self.requests_count = 0
        if not baseui:
            # set up environment for new ui
            extensions.loadall(self.ui)
            extensions.populateui(self.ui)

    def refresh(self):
        if self.ui:
            refreshinterval = self.ui.configint(b'web', b'refreshinterval')
        else:
            item = configitems.coreitems[b'web'][b'refreshinterval']
            refreshinterval = item.default

        # refreshinterval <= 0 means to always refresh.
        if (
            refreshinterval > 0
            and self.lastrefresh + refreshinterval > time.time()
        ):
            return

        if self.baseui:
            u = self.baseui.copy()
        else:
            u = uimod.ui.load()
            u.setconfig(b'ui', b'report_untrusted', b'off', b'hgwebdir')
            u.setconfig(b'ui', b'nontty', b'true', b'hgwebdir')
            # displaying bundling progress bar while serving feels wrong and may
            # break some wsgi implementations.
            u.setconfig(b'progress', b'disable', b'true', b'hgweb')

        if not isinstance(self.conf, (dict, list, tuple)):
            map = {b'paths': b'hgweb-paths'}
            if not os.path.exists(self.conf):
                raise error.Abort(_(b'config file %s not found!') % self.conf)
            u.readconfig(self.conf, remap=map, trust=True)
            paths = []
            for name, ignored in u.configitems(b'hgweb-paths'):
                for path in u.configlist(b'hgweb-paths', name):
                    paths.append((name, path))
        elif isinstance(self.conf, (list, tuple)):
            paths = self.conf
        elif isinstance(self.conf, dict):
            paths = self.conf.items()
        extensions.populateui(u)

        repos = findrepos(paths)
        for prefix, root in u.configitems(b'collections'):
            prefix = util.pconvert(prefix)
            for path in scmutil.walkrepos(root, followsym=True):
                repo = os.path.normpath(path)
                name = util.pconvert(repo)
                if name.startswith(prefix):
                    name = name[len(prefix) :]
                repos.append((name.lstrip(b'/'), repo))

        self.repos = repos
        self.ui = u
        self.gc_full_collect_rate = self.ui.configint(
            b'experimental', b'web.full-garbage-collection-rate'
        )
        self.gc_full_collections_done = 0
        encoding.encoding = self.ui.config(b'web', b'encoding')
        self.style = self.ui.config(b'web', b'style')
        self.templatepath = self.ui.config(
            b'web', b'templates', untrusted=False
        )
        self.stripecount = self.ui.config(b'web', b'stripes')
        if self.stripecount:
            self.stripecount = int(self.stripecount)
        prefix = self.ui.config(b'web', b'prefix')
        if prefix.startswith(b'/'):
            prefix = prefix[1:]
        if prefix.endswith(b'/'):
            prefix = prefix[:-1]
        self.prefix = prefix
        self.lastrefresh = time.time()

    def run(self):
        if not encoding.environ.get(b'GATEWAY_INTERFACE', b'').startswith(
            b"CGI/1."
        ):
            raise RuntimeError(
                b"This function is only intended to be "
                b"called while running as a CGI script."
            )
        wsgicgi.launch(self)

    def __call__(self, env, respond):
        baseurl = self.ui.config(b'web', b'baseurl')
        req = requestmod.parserequestfromenv(env, altbaseurl=baseurl)
        res = requestmod.wsgiresponse(req, respond)

        return self.run_wsgi(req, res)

    def run_wsgi(self, req, res):
        profile = self.ui.configbool(b'profiling', b'enabled')
        with profiling.profile(self.ui, enabled=profile):
            try:
                for r in self._runwsgi(req, res):
                    yield r
            finally:
                # There are known cycles in localrepository that prevent
                # those objects (and tons of held references) from being
                # collected through normal refcounting.
                # In some cases, the resulting memory consumption can
                # be tamed by performing explicit garbage collections.
                # In presence of actual leaks or big long-lived caches, the
                # impact on performance of such collections can become a
                # problem, hence the rate shouldn't be set too low.
                # See "Collecting the oldest generation" in
                # https://devguide.python.org/garbage_collector
                # for more about such trade-offs.
                rate = self.gc_full_collect_rate

                # this is not thread safe, but the consequence (skipping
                # a garbage collection) is arguably better than risking
                # to have several threads perform a collection in parallel
                # (long useless wait on all threads).
                self.requests_count += 1
                if rate > 0 and self.requests_count % rate == 0:
                    gc.collect()
                    self.gc_full_collections_done += 1
                else:
                    gc.collect(generation=1)

    def _runwsgi(self, req, res):
        try:
            self.refresh()

            csp, nonce = cspvalues(self.ui)
            if csp:
                res.headers[b'Content-Security-Policy'] = csp

            virtual = req.dispatchpath.strip(b'/')
            tmpl = self.templater(req, nonce)
            ctype = tmpl.render(b'mimetype', {b'encoding': encoding.encoding})

            # Global defaults. These can be overridden by any handler.
            res.status = b'200 Script output follows'
            res.headers[b'Content-Type'] = ctype

            # a static file
            if virtual.startswith(b'static/') or b'static' in req.qsparams:
                if virtual.startswith(b'static/'):
                    fname = virtual[7:]
                else:
                    fname = req.qsparams[b'static']
                static = self.ui.config(b"web", b"static", untrusted=False)
                staticfile(self.templatepath, static, fname, res)
                return res.sendresponse()

            # top-level index

            repos = dict(self.repos)

            if (not virtual or virtual == b'index') and virtual not in repos:
                return self.makeindex(req, res, tmpl)

            # nested indexes and hgwebs

            if virtual.endswith(b'/index') and virtual not in repos:
                subdir = virtual[: -len(b'index')]
                if any(r.startswith(subdir) for r in repos):
                    return self.makeindex(req, res, tmpl, subdir)

            def _virtualdirs():
                # Check the full virtual path, and each parent
                yield virtual
                for p in pathutil.finddirs(virtual):
                    yield p

            for virtualrepo in _virtualdirs():
                real = repos.get(virtualrepo)
                if real:
                    # Re-parse the WSGI environment to take into account our
                    # repository path component.
                    uenv = req.rawenv
                    if pycompat.ispy3:
                        uenv = {
                            k.decode('latin1'): v
                            for k, v in pycompat.iteritems(uenv)
                        }
                    req = requestmod.parserequestfromenv(
                        uenv,
                        reponame=virtualrepo,
                        altbaseurl=self.ui.config(b'web', b'baseurl'),
                        # Reuse wrapped body file object otherwise state
                        # tracking can get confused.
                        bodyfh=req.bodyfh,
                    )
                    try:
                        # ensure caller gets private copy of ui
                        repo = hg.repository(self.ui.copy(), real)
                        return hgweb_mod.hgweb(repo).run_wsgi(req, res)
                    except IOError as inst:
                        msg = encoding.strtolocal(inst.strerror)
                        raise ErrorResponse(HTTP_SERVER_ERROR, msg)
                    except error.RepoError as inst:
                        raise ErrorResponse(HTTP_SERVER_ERROR, bytes(inst))

            # browse subdirectories
            subdir = virtual + b'/'
            if [r for r in repos if r.startswith(subdir)]:
                return self.makeindex(req, res, tmpl, subdir)

            # prefixes not found
            res.status = b'404 Not Found'
            res.setbodygen(tmpl.generate(b'notfound', {b'repo': virtual}))
            return res.sendresponse()

        except ErrorResponse as e:
            res.status = statusmessage(e.code, pycompat.bytestr(e))
            res.setbodygen(
                tmpl.generate(b'error', {b'error': e.message or b''})
            )
            return res.sendresponse()
        finally:
            del tmpl

    def makeindex(self, req, res, tmpl, subdir=b""):
        self.refresh()
        sortable = [b"name", b"description", b"contact", b"lastchange"]
        sortcolumn, descending = None, False
        if b'sort' in req.qsparams:
            sortcolumn = req.qsparams[b'sort']
            descending = sortcolumn.startswith(b'-')
            if descending:
                sortcolumn = sortcolumn[1:]
            if sortcolumn not in sortable:
                sortcolumn = b""

        sort = [
            (
                b"sort_%s" % column,
                b"%s%s"
                % (
                    (not descending and column == sortcolumn) and b"-" or b"",
                    column,
                ),
            )
            for column in sortable
        ]

        self.refresh()

        entries = indexentries(
            self.ui,
            self.repos,
            req,
            self.stripecount,
            sortcolumn=sortcolumn,
            descending=descending,
            subdir=subdir,
        )

        mapping = {
            b'entries': entries,
            b'subdir': subdir,
            b'pathdef': hgweb_mod.makebreadcrumb(b'/' + subdir, self.prefix),
            b'sortcolumn': sortcolumn,
            b'descending': descending,
        }
        mapping.update(sort)
        res.setbodygen(tmpl.generate(b'index', mapping))
        return res.sendresponse()

    def templater(self, req, nonce):
        def config(*args, **kwargs):
            kwargs.setdefault('untrusted', True)
            return self.ui.config(*args, **kwargs)

        vars = {}
        styles, (style, mapfile, fp) = hgweb_mod.getstyle(
            req, config, self.templatepath
        )
        if style == styles[0]:
            vars[b'style'] = style

        sessionvars = webutil.sessionvars(vars, b'?')
        logourl = config(b'web', b'logourl')
        logoimg = config(b'web', b'logoimg')
        staticurl = (
            config(b'web', b'staticurl')
            or req.apppath.rstrip(b'/') + b'/static/'
        )
        if not staticurl.endswith(b'/'):
            staticurl += b'/'

        defaults = {
            b"encoding": encoding.encoding,
            b"url": req.apppath + b'/',
            b"logourl": logourl,
            b"logoimg": logoimg,
            b"staticurl": staticurl,
            b"sessionvars": sessionvars,
            b"style": style,
            b"nonce": nonce,
        }
        templatekeyword = registrar.templatekeyword(defaults)

        @templatekeyword(b'motd', requires=())
        def motd(context, mapping):
            if self.motd is not None:
                yield self.motd
            else:
                yield config(b'web', b'motd')

        return templater.templater.frommapfile(
            mapfile, fp=fp, defaults=defaults
        )