dirstate: avoid a race with multiple commits in the same process
(issue2264, issue2516)
The race happens when two commits in a row change the same file
without changing its size, *if* those two commits happen in the same
second in the same process while holding the same repo lock. For
example:
commit 1:
M a
M b
commit 2: # same process, same second, same repo lock
M b # modify b without changing its size
M c
This first manifested in transplant, which is the most common way to
do multiple commits in the same process. But it can manifest in any
script or extension that does multiple commits under the same repo
lock. (Thus, the test script tests both transplant and a custom script.)
The problem was that dirstate.status() failed to notice the change to
b when localrepo is about to do the second commit, meaning that change
gets left in the working directory. In the context of transplant, that
means either a crash ("RuntimeError: nothing committed after
transplant") or a silently inaccurate transplant, depending on whether
any other files were modified by the second transplanted changeset.
The fix is to make status() work a little harder when we have
previously marked files as clean (state 'normal') in the same process.
Specifically, dirstate.normal() adds files to self._lastnormal, and
other state-changing methods remove them. Then dirstate.status() puts
any files in self._lastnormal into state 'lookup', which will make
localrepository.status() read file contents to see if it has really
changed. So we pay a small performance penalty for the second (and
subsequent) commits in the same process, without affecting the common
case. Anything that does lots of status updates and checks in the
same process could suffer a performance hit.
Incidentally, there is a simpler fix: call dirstate.normallookup() on
every file updated by commit() at the end of the commit. The trouble
with that solution is that it imposes a performance penalty on the
common case: it means the next status-dependent hg command after every
"hg commit" will be a little bit slower. The patch here is more
complex, but only affects performance for the uncommon case.
# ancestor.py - generic DAG ancestor algorithm for mercurial
#
# Copyright 2006 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.
import heapq
def ancestor(a, b, pfunc):
"""
Returns the common ancestor of a and b that is furthest from a
root (as measured by longest path) or None if no ancestor is
found. If there are multiple common ancestors at the same
distance, the first one found is returned.
pfunc must return a list of parent vertices for a given vertex
"""
if a == b:
return a
a, b = sorted([a, b])
# find depth from root of all ancestors
# depth is stored as a negative for heapq
parentcache = {}
visit = [a, b]
depth = {}
while visit:
vertex = visit[-1]
pl = pfunc(vertex)
parentcache[vertex] = pl
if not pl:
depth[vertex] = 0
visit.pop()
else:
for p in pl:
if p == a or p == b: # did we find a or b as a parent?
return p # we're done
if p not in depth:
visit.append(p)
if visit[-1] == vertex:
# -(maximum distance of parents + 1)
depth[vertex] = min([depth[p] for p in pl]) - 1
visit.pop()
# traverse ancestors in order of decreasing distance from root
def ancestors(vertex):
h = [(depth[vertex], vertex)]
seen = set()
while h:
d, n = heapq.heappop(h)
if n not in seen:
seen.add(n)
yield (d, n)
for p in parentcache[n]:
heapq.heappush(h, (depth[p], p))
def generations(vertex):
sg, s = None, set()
for g, v in ancestors(vertex):
if g != sg:
if sg:
yield sg, s
sg, s = g, set((v,))
else:
s.add(v)
yield sg, s
x = generations(a)
y = generations(b)
gx = x.next()
gy = y.next()
# increment each ancestor list until it is closer to root than
# the other, or they match
try:
while 1:
if gx[0] == gy[0]:
for v in gx[1]:
if v in gy[1]:
return v
gy = y.next()
gx = x.next()
elif gx[0] > gy[0]:
gy = y.next()
else:
gx = x.next()
except StopIteration:
return None