bisect: switch to rev-based calculation
- use revlog.parentrevs in search loops
- calculate stop directly in O(N) without reachable
- move badrev check into candidates
# bisect extension for mercurial
#
# Copyright 2005, 2006 Benoit Boissinot <benoit.boissinot@ens-lyon.org>
# Inspired by git bisect, extension skeleton taken from mq.py.
#
# This software may be used and distributed according to the terms
# of the GNU General Public License, incorporated herein by reference.
from mercurial.i18n import _
from mercurial import hg, util, commands, cmdutil
import os, sys
class bisect(object):
"""dichotomic search in the DAG of changesets"""
def __init__(self, ui, repo):
self.repo = repo
self.path = repo.join("bisect")
self.opener = util.opener(self.path)
self.ui = ui
self.goodnodes = []
self.badnode = None
self.good_path = "good"
self.bad_path = "bad"
self.is_reset = False
if os.path.exists(os.path.join(self.path, self.good_path)):
self.goodnodes = self.opener(self.good_path).read().splitlines()
self.goodnodes = [hg.bin(x) for x in self.goodnodes]
if os.path.exists(os.path.join(self.path, self.bad_path)):
r = self.opener(self.bad_path).read().splitlines()
if r:
self.badnode = hg.bin(r.pop(0))
def write(self):
if self.is_reset:
return
if not os.path.isdir(self.path):
os.mkdir(self.path)
f = self.opener(self.good_path, "w")
f.write("\n".join([hg.hex(r) for r in self.goodnodes]))
if len(self.goodnodes) > 0:
f.write("\n")
f = self.opener(self.bad_path, "w")
if self.badnode:
f.write(hg.hex(self.badnode) + "\n")
def init(self):
"""start a new bisection"""
if os.path.isdir(self.path):
raise util.Abort(_("bisect directory already exists\n"))
os.mkdir(self.path)
return 0
def reset(self):
"""finish a bisection"""
if os.path.isdir(self.path):
sl = [os.path.join(self.path, p)
for p in [self.bad_path, self.good_path]]
for s in sl:
if os.path.exists(s):
os.unlink(s)
os.rmdir(self.path)
# Not sure about this
#self.ui.write("Going back to tip\n")
#self.repo.update(self.repo.changelog.tip())
self.is_reset = True
return 0
def candidates(self):
"""
returns (anc, n_child) where anc is the set of the ancestors of head
and n_child is a dictionary with the following mapping:
node -> number of ancestors (self included)
ancestors of goodnodes are used as lower limit.
"""
cl = self.repo.changelog
clparents = cl.parentrevs
bad = self.badnode
badrev = cl.rev(bad)
# add goodrevs and all ancestors to stop
stop = dict.fromkeys([cl.rev(n) for n in self.goodnodes])
for rev in xrange(cl.count(), -1, -1):
if rev in stop:
for prev in clparents(rev):
stop[prev] = None
if badrev in stop:
raise util.Abort(_("Inconsistent state, %s:%s is good and bad")
% (badrev, hg.short(bad)))
# build a dict of node -> [number of children, {}]
n_child = {badrev: [0, {}]}
for rev in xrange(badrev + 1):
if rev not in n_child:
n_child[rev] = [0, {}]
# increment count of each parent
for p in clparents(rev):
if p != hg.nullrev:
n_child[p][0] += 1
for rev in xrange(badrev + 1):
for p in clparents(rev):
if p != hg.nullrev:
n_child[p][0] -= 1
if not rev in stop:
n_child[rev][1].update(n_child[p][1])
if n_child[p][0] == 0:
n_child[p] = len(n_child[p][1])
if not rev in stop:
n_child[rev][1][rev] = 1
if n_child[rev][0] == 0:
if rev == badrev:
anc = n_child[rev][1]
n_child[rev] = len(n_child[rev][1])
if badrev not in anc:
raise util.Abort(_("Could not find the first bad revision"))
return anc, n_child
def next(self):
if not self.badnode:
raise util.Abort(_("You should give at least one bad revision"))
if not self.goodnodes:
self.ui.warn(_("No good revision given\n"))
self.ui.warn(_("Marking the first revision as good\n"))
ancestors, n_child = self.candidates()
# have we narrowed it down to one entry?
tot = len(ancestors)
if tot == 1:
self.ui.write(_("The first bad revision is:\n"))
displayer = cmdutil.show_changeset(self.ui, self.repo, {})
displayer.show(changenode=self.badnode)
return None
# find the best node to test
best_rev = None
best_len = -1
for n in ancestors:
a = n_child[n] # number of children
b = tot - a # number of ancestors
value = min(a, b) # how good is this test?
if value > best_len:
best_len = value
best_rev = n
assert best_rev is not None
best_node = self.repo.changelog.node(best_rev)
# compute the approximate number of remaining tests
nb_tests = 0
q, r = divmod(tot, 2)
while q:
nb_tests += 1
q, r = divmod(q, 2)
self.ui.write(_("Testing changeset %s:%s "
"(%s changesets remaining, ~%s tests)\n")
% (best_rev, hg.short(best_node), tot, nb_tests))
return best_node
def autonext(self):
"""find and update to the next revision to test"""
node = self.next()
if node is not None:
cmdutil.bail_if_changed(self.repo)
return hg.clean(self.repo, node)
def autogood(self, rev=None):
"""mark revision as good and update to the next revision to test"""
self.goodnodes.append(self.repo.lookup(rev or '.'))
if self.badnode:
return self.autonext()
def autobad(self, rev=None):
"""mark revision as bad and update to the next revision to test"""
self.badnode = self.repo.lookup(rev or '.')
if self.goodnodes:
self.autonext()
# should we put it in the class ?
def test(ui, repo, rev):
"""test the bisection code"""
b = bisect(ui, repo)
node = repo.lookup(rev)
ui.write("testing with rev %s\n" % hg.hex(node))
anc = b.ancestors()
while len(anc) > 1:
if not node in anc:
ui.warn("failure while bisecting\n")
sys.exit(1)
ui.write("it worked :)\n")
new_node = b.next()
ui.write("choosing if good or bad\n")
if node in b.ancestors(head=new_node):
b.bad(new_node)
ui.write("it is bad\n")
else:
b.good(new_node)
ui.write("it is good\n")
anc = b.ancestors()
#repo.update(new_node, force=True)
for v in anc:
if v != node:
ui.warn("fail to found cset! :(\n")
return 1
ui.write("Found bad cset: %s\n" % hg.hex(b.badnode))
ui.write("Everything is ok :)\n")
return 0
def bisect_run(ui, repo, cmd=None, *args):
"""Dichotomic search in the DAG of changesets
This extension helps to find changesets which cause problems.
To use, mark the earliest changeset you know introduces the problem
as bad, then mark the latest changeset which is free from the problem
as good. Bisect will update your working directory to a revision for
testing. Once you have performed tests, mark the working directory
as bad or good and bisect will either update to another candidate
changeset or announce that it has found the bad revision.
Note: bisect expects bad revisions to be descendants of good revisions.
If you are looking for the point at which a problem was fixed, then make
the problem-free state "bad" and the problematic state "good."
For subcommands see "hg bisect help\"
"""
def help_(cmd=None, *args):
"""show help for a given bisect subcommand or all subcommands"""
cmdtable = bisectcmdtable
if cmd:
doc = cmdtable[cmd][0].__doc__
synopsis = cmdtable[cmd][2]
ui.write(synopsis + "\n")
ui.write("\n" + doc + "\n")
return
ui.write(_("list of subcommands for the bisect extension\n\n"))
cmds = cmdtable.keys()
cmds.sort()
m = max([len(c) for c in cmds])
for cmd in cmds:
doc = cmdtable[cmd][0].__doc__.splitlines(0)[0].rstrip()
ui.write(" %-*s %s\n" % (m, cmd, doc))
b = bisect(ui, repo)
bisectcmdtable = {
"init": (b.init, 0, _("hg bisect init")),
"bad": (b.autobad, 1, _("hg bisect bad [<rev>]")),
"good": (b.autogood, 1, _("hg bisect good [<rev>]")),
"next": (b.autonext, 0, _("hg bisect next")),
"reset": (b.reset, 0, _("hg bisect reset")),
"help": (help_, 1, _("hg bisect help [<subcommand>]")),
}
if not bisectcmdtable.has_key(cmd):
ui.warn(_("bisect: Unknown sub-command\n"))
return help_()
if len(args) > bisectcmdtable[cmd][1]:
ui.warn(_("bisect: Too many arguments\n"))
return help_()
ret = bisectcmdtable[cmd][0](*args)
b.write()
return ret
cmdtable = {
"bisect": (bisect_run, [], _("hg bisect [help|init|reset|next|good|bad]")),
#"bisect-test": (test, [], "hg bisect-test rev"),
}