##// END OF EJS Templates
changegroup: remove reordering control (BC)...
changegroup: remove reordering control (BC) This logic - including the experimental bundle.reorder option - was originally added in a8e3931e3fb5 in 2011 and then later ported to changegroup.py. The intent of this option and associated logic is to control the ordering of revisions in deltagroups in changegroups. At the time it was implemented, only changegroup version 1 existed and generaldelta revlogs were just coming into the world. Changegroup version 1 requires that deltas be made against the last revision sent over the wire. Used with generaldelta, this created an impedance mismatch of sorts and resulted in changegroup producers spending a lot of time recomputing deltas. Revision reordering was introduced so outgoing revisions would be sent in "generaldelta order" and producers would be able to reuse internal deltas from storage. Later on, we introduced changegroup version 2. It supported denoting which revision a delta was against. So we no longer needed to sort outgoing revisions to ensure optimal delta generation from the producer. So, subsequent changegroup versions disabled reordering. We also later made the changelog not store deltas by default. And we also made the changelog send out deltas in storage order. Why we do this for changelog, I'm not sure. Maybe we want to preserve revision order across clones? It doesn't really matter for this commit. Fast forward to 2018. We want to abstract storage backends. And having changegroup code require knowledge about how deltas are stored internally interferes with that goal. This commit removes reordering control from changegroup generation. After this commit, the reordering behavior is: * The changelog is always sent out in storage order (no behavior change). * Non-changelog generaldelta revlogs are reordered to always be in DAG topological order (previously, generaldelta revlogs would be emitted in storage order for version 2 and 3 changegroups). * Non-changelog non-generaldelta revlogs are sent in storage order (no behavior change). * There exists no config option to override behavior. The big difference here is that generaldelta revlogs now *always* have their revisions sorted in DAG order before going out over the wire. This behavior was previously only done for changegroup version 1. Version 2 and version 3 changegroups disabled reordering because the interchange format supported encoding arbitrary delta parents, so reordering wasn't strictly necessary. I can think of a few significant implications for this change. Because changegroup receivers will now see non-changelog revisions in DAG order instead of storage order, the internal storage order of manifests and files may differ substantially between producer and consumer. I don't think this matters that much, since the storage order of manifests and files is largely hidden from users. Only the storage order of changelog matters (because `hg log` shows the changelog in storage order). I don't think there should be any controversy here. The reordering of revisions has implications for changegroup producers. Previously, generaldelta revlogs would be emitted in storage order. And in the common case, the internally-stored delta could effectively be copied from disk into the deltagroup delta. This meant that emitting delta groups for generaldelta revlogs would be mostly linear read I/O. This is desirable for performance. With us now reordering generaldelta revlog revisions in DAG order, the read operations may use more random I/O instead of sequential I/O. This could result in performance loss. But with the prevalence of SSDs and fast random I/O, I'm not too worried. (Note: the optimal emission order for revlogs is actually delta encoding order. But the changegroup code wasn't doing that before or after this change. We could potentially implement that in a later commit.) Changegroups in DAG order will have implications for receivers. Previously, receiving storage order might mean seeing a number of interleaved branches. This would mean long delta chains, sparse I/O, and possibly more fulltext revisions instead of deltas, blowing up storage storage. (This is the same set of problems that sparse revlogs aims to address.) With the producer now sending revisions in DAG order, the receiver also stores revisions in DAG order. That means revisions for the same DAG branch are all grouped together. And this should yield better storage outcomes. In other words, sending the reordered changegroup allows the receiver to have better storage order and for the producer to not propagate its (possibly sub-optimal) internal storage order. On the mozilla-unified repository, this change influences bundle generation: $ hg bundle -t none-v2 -a before: time: real 355.680 secs (user 256.790+0.000 sys 16.820+0.000) after: time: real 382.950 secs (user 281.700+0.000 sys 17.690+0.000) before: 7,150,228,967 bytes (uncompressed) after: 7,041,556,273 bytes (uncompressed) before: 1,669,063,234 bytes (zstd l=3) after: 1,628,598,830 bytes (zstd l=3) $ hg unbundle before: time: real 511.910 secs (user 466.750+0.000 sys 32.680+0.000) after: time: real 487.790 secs (user 443.940+0.000 sys 30.840+0.000) 00manifest.d size: source: 274,924,292 bytes before: 304,741,626 bytes after: 245,252,087 bytes .hg/store total file size: source: 2,649,133,490 before: 2,680,888,130 after: 2,627,875,673 We see the bundle size drop. That's probably because if a revlog internally isn't storing a delta, it will choose to delta against the last emitted revision. And on repos with interleaved branches (like mozilla-unified), the previous revision could be an unrelated branch and therefore be a large delta. But with this patch, the previous revision is likely p1 or p2 and a delta should be small. We also see the manifest size drop by ~50 MB. It's worth noting that the manifest actually *increased* in size by ~25 MB in the old strategy and decreased ~25 MB from its source in the new strategy. Again, my explanation for this is that the DAG ordering in the changegroup is resulting in better grouping of revisions in the receiver, which results in more compact delta chains and higher storage efficiency. Unbundle time also dropped. I suspect this is due to the revlog having to work less to compute deltas since the incoming deltas are more optimal. i.e. the receiver spends less time resolving fulltext revisions as incoming deltas bounce around between DAG branches and delta chains. We also see bundle generation time increase. This is not desirable. However, the regression is only significant on the original repository: if we generate a bundle from the repository created from the new, always reordered bundles, we're close to baseline (if not at it with expected noise): $ hg bundle -t none-v2 -a before (original): time: real 355.680 secs (user 256.790+0.000 sys 16.820+0.000) after (original): time: real 382.950 secs (user 281.700+0.000 sys 17.690+0.000) after (new repo): time: real 362.280 secs (user 260.300+0.000 sys 17.700+0.000) This regression is a bit worrying because it will impact serving canonical repositories (that don't have optimal internal storage unless they are reordered - possibly as part of running `hg debugupgraderepo`). However, this regression will only be noticed by very large changegroups. And I'm guessing/hoping that any repository that large is using clonebundles to mitigate server load. Again, sending DAG order isn't the optimal send order for servers: sending in storage-delta order is. But in order to enable storage-optimal send order, we'll need a storage API that handles sorting. Future commits will introduce such an API. Differential Revision: https://phab.mercurial-scm.org/D4721

File last commit:

r39813:b63dee7b default
r39897:db5501d9 default
Show More
hgweb_mod.py
475 lines | 17.6 KiB | text/x-python | PythonLexer
# hgweb/hgweb_mod.py - Web interface for a repository.
#
# Copyright 21 May 2005 - (c) 2005 Jake Edge <jake@edge2.net>
# Copyright 2005-2007 Matt Mackall <mpm@selenic.com>
#
# 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 contextlib
import os
from .common import (
ErrorResponse,
HTTP_BAD_REQUEST,
cspvalues,
permhooks,
statusmessage,
)
from .. import (
encoding,
error,
formatter,
hg,
hook,
profiling,
pycompat,
registrar,
repoview,
templatefilters,
templater,
templateutil,
ui as uimod,
util,
wireprotoserver,
)
from . import (
request as requestmod,
webcommands,
webutil,
wsgicgi,
)
def getstyle(req, configfn, templatepath):
styles = (
req.qsparams.get('style', None),
configfn('web', 'style'),
'paper',
)
return styles, templater.stylemap(styles, templatepath)
def makebreadcrumb(url, prefix=''):
'''Return a 'URL breadcrumb' list
A 'URL breadcrumb' is a list of URL-name pairs,
corresponding to each of the path items on a URL.
This can be used to create path navigation entries.
'''
if url.endswith('/'):
url = url[:-1]
if prefix:
url = '/' + prefix + url
relpath = url
if relpath.startswith('/'):
relpath = relpath[1:]
breadcrumb = []
urlel = url
pathitems = [''] + relpath.split('/')
for pathel in reversed(pathitems):
if not pathel or not urlel:
break
breadcrumb.append({'url': urlel, 'name': pathel})
urlel = os.path.dirname(urlel)
return templateutil.mappinglist(reversed(breadcrumb))
class requestcontext(object):
"""Holds state/context for an individual request.
Servers can be multi-threaded. Holding state on the WSGI application
is prone to race conditions. Instances of this class exist to hold
mutable and race-free state for requests.
"""
def __init__(self, app, repo, req, res):
self.repo = repo
self.reponame = app.reponame
self.req = req
self.res = res
self.maxchanges = self.configint('web', 'maxchanges')
self.stripecount = self.configint('web', 'stripes')
self.maxshortchanges = self.configint('web', 'maxshortchanges')
self.maxfiles = self.configint('web', 'maxfiles')
self.allowpull = self.configbool('web', 'allow-pull')
# we use untrusted=False to prevent a repo owner from using
# web.templates in .hg/hgrc to get access to any file readable
# by the user running the CGI script
self.templatepath = self.config('web', 'templates', untrusted=False)
# This object is more expensive to build than simple config values.
# It is shared across requests. The app will replace the object
# if it is updated. Since this is a reference and nothing should
# modify the underlying object, it should be constant for the lifetime
# of the request.
self.websubtable = app.websubtable
self.csp, self.nonce = cspvalues(self.repo.ui)
# Trust the settings from the .hg/hgrc files by default.
def config(self, section, name, default=uimod._unset, untrusted=True):
return self.repo.ui.config(section, name, default,
untrusted=untrusted)
def configbool(self, section, name, default=uimod._unset, untrusted=True):
return self.repo.ui.configbool(section, name, default,
untrusted=untrusted)
def configint(self, section, name, default=uimod._unset, untrusted=True):
return self.repo.ui.configint(section, name, default,
untrusted=untrusted)
def configlist(self, section, name, default=uimod._unset, untrusted=True):
return self.repo.ui.configlist(section, name, default,
untrusted=untrusted)
def archivelist(self, nodeid):
return webutil.archivelist(self.repo.ui, nodeid)
def templater(self, req):
# determine scheme, port and server name
# this is needed to create absolute urls
logourl = self.config('web', 'logourl')
logoimg = self.config('web', 'logoimg')
staticurl = (self.config('web', 'staticurl')
or req.apppath.rstrip('/') + '/static/')
if not staticurl.endswith('/'):
staticurl += '/'
# figure out which style to use
vars = {}
styles, (style, mapfile) = getstyle(req, self.config,
self.templatepath)
if style == styles[0]:
vars['style'] = style
sessionvars = webutil.sessionvars(vars, '?')
if not self.reponame:
self.reponame = (self.config('web', 'name', '')
or req.reponame
or req.apppath
or self.repo.root)
filters = {}
templatefilter = registrar.templatefilter(filters)
@templatefilter('websub', intype=bytes)
def websubfilter(text):
return templatefilters.websub(text, self.websubtable)
# create the templater
# TODO: export all keywords: defaults = templatekw.keywords.copy()
defaults = {
'url': req.apppath + '/',
'logourl': logourl,
'logoimg': logoimg,
'staticurl': staticurl,
'urlbase': req.advertisedbaseurl,
'repo': self.reponame,
'encoding': encoding.encoding,
'sessionvars': sessionvars,
'pathdef': makebreadcrumb(req.apppath),
'style': style,
'nonce': self.nonce,
}
templatekeyword = registrar.templatekeyword(defaults)
@templatekeyword('motd', requires=())
def motd(context, mapping):
yield self.config('web', 'motd')
tres = formatter.templateresources(self.repo.ui, self.repo)
tmpl = templater.templater.frommapfile(mapfile,
filters=filters,
defaults=defaults,
resources=tres)
return tmpl
def sendtemplate(self, name, **kwargs):
"""Helper function to send a response generated from a template."""
kwargs = pycompat.byteskwargs(kwargs)
self.res.setbodygen(self.tmpl.generate(name, kwargs))
return self.res.sendresponse()
class hgweb(object):
"""HTTP server for individual repositories.
Instances of this class serve HTTP responses for a particular
repository.
Instances are typically used as WSGI applications.
Some servers are multi-threaded. On these servers, there may
be multiple active threads inside __call__.
"""
def __init__(self, repo, name=None, baseui=None):
if isinstance(repo, bytes):
if baseui:
u = baseui.copy()
else:
u = uimod.ui.load()
r = hg.repository(u, repo)
else:
# we trust caller to give us a private copy
r = repo
r.ui.setconfig('ui', 'report_untrusted', 'off', 'hgweb')
r.baseui.setconfig('ui', 'report_untrusted', 'off', 'hgweb')
r.ui.setconfig('ui', 'nontty', 'true', 'hgweb')
r.baseui.setconfig('ui', 'nontty', 'true', 'hgweb')
# resolve file patterns relative to repo root
r.ui.setconfig('ui', 'forcecwd', r.root, 'hgweb')
r.baseui.setconfig('ui', 'forcecwd', r.root, 'hgweb')
# it's unlikely that we can replace signal handlers in WSGI server,
# and mod_wsgi issues a big warning. a plain hgweb process (with no
# threading) could replace signal handlers, but we don't bother
# conditionally enabling it.
r.ui.setconfig('ui', 'signal-safe-lock', 'false', 'hgweb')
r.baseui.setconfig('ui', 'signal-safe-lock', 'false', 'hgweb')
# displaying bundling progress bar while serving feel wrong and may
# break some wsgi implementation.
r.ui.setconfig('progress', 'disable', 'true', 'hgweb')
r.baseui.setconfig('progress', 'disable', 'true', 'hgweb')
self._repos = [hg.cachedlocalrepo(self._webifyrepo(r))]
self._lastrepo = self._repos[0]
hook.redirect(True)
self.reponame = name
def _webifyrepo(self, repo):
repo = getwebview(repo)
self.websubtable = webutil.getwebsubs(repo)
return repo
@contextlib.contextmanager
def _obtainrepo(self):
"""Obtain a repo unique to the caller.
Internally we maintain a stack of cachedlocalrepo instances
to be handed out. If one is available, we pop it and return it,
ensuring it is up to date in the process. If one is not available,
we clone the most recently used repo instance and return it.
It is currently possible for the stack to grow without bounds
if the server allows infinite threads. However, servers should
have a thread limit, thus establishing our limit.
"""
if self._repos:
cached = self._repos.pop()
r, created = cached.fetch()
else:
cached = self._lastrepo.copy()
r, created = cached.fetch()
if created:
r = self._webifyrepo(r)
self._lastrepo = cached
self.mtime = cached.mtime
try:
yield r
finally:
self._repos.append(cached)
def run(self):
"""Start a server from CGI environment.
Modern servers should be using WSGI and should avoid this
method, if possible.
"""
if not encoding.environ.get('GATEWAY_INTERFACE',
'').startswith("CGI/1."):
raise RuntimeError("This function is only intended to be "
"called while running as a CGI script.")
wsgicgi.launch(self)
def __call__(self, env, respond):
"""Run the WSGI application.
This may be called by multiple threads.
"""
req = requestmod.parserequestfromenv(env)
res = requestmod.wsgiresponse(req, respond)
return self.run_wsgi(req, res)
def run_wsgi(self, req, res):
"""Internal method to run the WSGI application.
This is typically only called by Mercurial. External consumers
should be using instances of this class as the WSGI application.
"""
with self._obtainrepo() as repo:
profile = repo.ui.configbool('profiling', 'enabled')
with profiling.profile(repo.ui, enabled=profile):
for r in self._runwsgi(req, res, repo):
yield r
def _runwsgi(self, req, res, repo):
rctx = requestcontext(self, repo, req, res)
# This state is global across all threads.
encoding.encoding = rctx.config('web', 'encoding')
rctx.repo.ui.environ = req.rawenv
if rctx.csp:
# hgwebdir may have added CSP header. Since we generate our own,
# replace it.
res.headers['Content-Security-Policy'] = rctx.csp
# /api/* is reserved for various API implementations. Dispatch
# accordingly. But URL paths can conflict with subrepos and virtual
# repos in hgwebdir. So until we have a workaround for this, only
# expose the URLs if the feature is enabled.
apienabled = rctx.repo.ui.configbool('experimental', 'web.apiserver')
if apienabled and req.dispatchparts and req.dispatchparts[0] == b'api':
wireprotoserver.handlewsgiapirequest(rctx, req, res,
self.check_perm)
return res.sendresponse()
handled = wireprotoserver.handlewsgirequest(
rctx, req, res, self.check_perm)
if handled:
return res.sendresponse()
# Old implementations of hgweb supported dispatching the request via
# the initial query string parameter instead of using PATH_INFO.
# If PATH_INFO is present (signaled by ``req.dispatchpath`` having
# a value), we use it. Otherwise fall back to the query string.
if req.dispatchpath is not None:
query = req.dispatchpath
else:
query = req.querystring.partition('&')[0].partition(';')[0]
# translate user-visible url structure to internal structure
args = query.split('/', 2)
if 'cmd' not in req.qsparams and args and args[0]:
cmd = args.pop(0)
style = cmd.rfind('-')
if style != -1:
req.qsparams['style'] = cmd[:style]
cmd = cmd[style + 1:]
# avoid accepting e.g. style parameter as command
if util.safehasattr(webcommands, cmd):
req.qsparams['cmd'] = cmd
if cmd == 'static':
req.qsparams['file'] = '/'.join(args)
else:
if args and args[0]:
node = args.pop(0).replace('%2F', '/')
req.qsparams['node'] = node
if args:
if 'file' in req.qsparams:
del req.qsparams['file']
for a in args:
req.qsparams.add('file', a)
ua = req.headers.get('User-Agent', '')
if cmd == 'rev' and 'mercurial' in ua:
req.qsparams['style'] = 'raw'
if cmd == 'archive':
fn = req.qsparams['node']
for type_, spec in webutil.archivespecs.iteritems():
ext = spec[2]
if fn.endswith(ext):
req.qsparams['node'] = fn[:-len(ext)]
req.qsparams['type'] = type_
else:
cmd = req.qsparams.get('cmd', '')
# process the web interface request
try:
rctx.tmpl = rctx.templater(req)
ctype = rctx.tmpl.render('mimetype',
{'encoding': encoding.encoding})
# check read permissions non-static content
if cmd != 'static':
self.check_perm(rctx, req, None)
if cmd == '':
req.qsparams['cmd'] = rctx.tmpl.render('default', {})
cmd = req.qsparams['cmd']
# Don't enable caching if using a CSP nonce because then it wouldn't
# be a nonce.
if rctx.configbool('web', 'cache') and not rctx.nonce:
tag = 'W/"%d"' % self.mtime
if req.headers.get('If-None-Match') == tag:
res.status = '304 Not Modified'
# Content-Type may be defined globally. It isn't valid on a
# 304, so discard it.
try:
del res.headers[b'Content-Type']
except KeyError:
pass
# Response body not allowed on 304.
res.setbodybytes('')
return res.sendresponse()
res.headers['ETag'] = tag
if cmd not in webcommands.__all__:
msg = 'no such method: %s' % cmd
raise ErrorResponse(HTTP_BAD_REQUEST, msg)
else:
# Set some globals appropriate for web handlers. Commands can
# override easily enough.
res.status = '200 Script output follows'
res.headers['Content-Type'] = ctype
return getattr(webcommands, cmd)(rctx)
except (error.LookupError, error.RepoLookupError) as err:
msg = pycompat.bytestr(err)
if (util.safehasattr(err, 'name') and
not isinstance(err, error.ManifestLookupError)):
msg = 'revision not found: %s' % err.name
res.status = '404 Not Found'
res.headers['Content-Type'] = ctype
return rctx.sendtemplate('error', error=msg)
except (error.RepoError, error.StorageError) as e:
res.status = '500 Internal Server Error'
res.headers['Content-Type'] = ctype
return rctx.sendtemplate('error', error=pycompat.bytestr(e))
except error.Abort as e:
res.status = '403 Forbidden'
res.headers['Content-Type'] = ctype
return rctx.sendtemplate('error', error=pycompat.bytestr(e))
except ErrorResponse as e:
for k, v in e.headers:
res.headers[k] = v
res.status = statusmessage(e.code, pycompat.bytestr(e))
res.headers['Content-Type'] = ctype
return rctx.sendtemplate('error', error=pycompat.bytestr(e))
def check_perm(self, rctx, req, op):
for permhook in permhooks:
permhook(rctx, req, op)
def getwebview(repo):
"""The 'web.view' config controls changeset filter to hgweb. Possible
values are ``served``, ``visible`` and ``all``. Default is ``served``.
The ``served`` filter only shows changesets that can be pulled from the
hgweb instance. The``visible`` filter includes secret changesets but
still excludes "hidden" one.
See the repoview module for details.
The option has been around undocumented since Mercurial 2.5, but no
user ever asked about it. So we better keep it undocumented for now."""
# experimental config: web.view
viewconfig = repo.ui.config('web', 'view', untrusted=True)
if viewconfig == 'all':
return repo.unfiltered()
elif viewconfig in repoview.filtertable:
return repo.filtered(viewconfig)
else:
return repo.filtered('served')