##// END OF EJS Templates
procutil: avoid using os.fork() to implement runbgcommand...
procutil: avoid using os.fork() to implement runbgcommand We ran into the following deadlock: - some command creates an ssh peer, then raises without explicitly closing the peer (hg id + extension in our case) - dispatch catches the exception, calls ui.log('commandfinish', ..) (the sshpeer is still not closed), which calls logtoprocess, which calls procutil.runbgcommand. - in the child of runbgcommand's fork(), between the fork and the exec, the opening of file descriptors triggers a gc which runs the destructor for sshpeer, which waits on ssh's stderr being closed, which never happens since ssh's stderr is held open by the parent of the fork where said destructor hasn't run Remotefilelog appears to have a hack around this deadlock as well. I don't know if there's more subtlety to it, because even though the problem is determistic, it is very fragile, so I didn't manage to reduce it. I can imagine three ways of tackling this problem: 1. don't run any python between fork and exec in runbgcommand 2. make the finalizer harmless after the fork 3. close the peer without relying on gc behavior This commit goes with 1, as forking without exec'ing is tricky in general in a language with gc finalizers. And maybe it's better in the presence of rust threads. A future commit will try 2 or 3. Performance wise: at low memory usage, it's an improvement. At higher memory usage, it's about 2x faster than before when ensurestart=True, but 2x slower when ensurestart=False. Not sure if that matters. The reason for that last bit is that the subprocess.Popen always waits for the execve to finish, and at high memory usage, execve is slow because it deallocates the large page table. Numbers and script: before after mem=1.0GB, ensurestart=True 52.1ms 26.0ms mem=1.0GB, ensurestart=False 14.7ms 26.0ms mem=0.5GB, ensurestart=True 23.2ms 11.2ms mem=0.5GB, ensurestart=False 6.2ms 11.3ms mem=0.2GB, ensurestart=True 15.7ms 7.4ms mem=0.2GB, ensurestart=False 4.3ms 8.1ms mem=0.0GB, ensurestart=True 2.3ms 0.7ms mem=0.0GB, ensurestart=False 0.8ms 0.8ms import time for memsize in [1_000_000_000, 500_000_000, 250_000_000, 0]: mem = 'a' * memsize for ensurestart in [True, False]: now = time.time() n = 100 for i in range(n): procutil.runbgcommand([b'true'], {}, ensurestart=ensurestart) after = time.time() ms = (after - now) / float(n) * 1000 print(f'mem={memsize / 1e9:.1f}GB, ensurestart={ensurestart} -> {ms:.1f}ms') Differential Revision: https://phab.mercurial-scm.org/D9019

File last commit:

r47619:1dc86c2a default
r47651:8759e22f default
Show More
parsers.py
352 lines | 11.2 KiB | text/x-python | PythonLexer
# parsers.py - Python implementation of parsers.c
#
# Copyright 2009 Olivia Mackall <olivia@selenic.com> and others
#
# 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 struct
import zlib
from ..node import nullid, nullrev
from .. import (
pycompat,
util,
)
from ..revlogutils import nodemap as nodemaputil
from ..revlogutils import constants as revlog_constants
stringio = pycompat.bytesio
_pack = struct.pack
_unpack = struct.unpack
_compress = zlib.compress
_decompress = zlib.decompress
# Some code below makes tuples directly because it's more convenient. However,
# code outside this module should always use dirstatetuple.
def dirstatetuple(*x):
# x is a tuple
return x
def gettype(q):
return int(q & 0xFFFF)
def offset_type(offset, type):
return int(int(offset) << 16 | type)
class BaseIndexObject(object):
# Format of an index entry according to Python's `struct` language
index_format = revlog_constants.INDEX_ENTRY_V1
# Size of a C unsigned long long int, platform independent
big_int_size = struct.calcsize(b'>Q')
# Size of a C long int, platform independent
int_size = struct.calcsize(b'>i')
# Size of the entire index format
index_size = revlog_constants.INDEX_ENTRY_V1.size
# An empty index entry, used as a default value to be overridden, or nullrev
null_item = (0, 0, 0, -1, -1, -1, -1, nullid)
@property
def nodemap(self):
msg = b"index.nodemap is deprecated, use index.[has_node|rev|get_rev]"
util.nouideprecwarn(msg, b'5.3', stacklevel=2)
return self._nodemap
@util.propertycache
def _nodemap(self):
nodemap = nodemaputil.NodeMap({nullid: nullrev})
for r in range(0, len(self)):
n = self[r][7]
nodemap[n] = r
return nodemap
def has_node(self, node):
"""return True if the node exist in the index"""
return node in self._nodemap
def rev(self, node):
"""return a revision for a node
If the node is unknown, raise a RevlogError"""
return self._nodemap[node]
def get_rev(self, node):
"""return a revision for a node
If the node is unknown, return None"""
return self._nodemap.get(node)
def _stripnodes(self, start):
if '_nodemap' in vars(self):
for r in range(start, len(self)):
n = self[r][7]
del self._nodemap[n]
def clearcaches(self):
self.__dict__.pop('_nodemap', None)
def __len__(self):
return self._lgt + len(self._extra)
def append(self, tup):
if '_nodemap' in vars(self):
self._nodemap[tup[7]] = len(self)
data = self.index_format.pack(*tup)
self._extra.append(data)
def _check_index(self, i):
if not isinstance(i, int):
raise TypeError(b"expecting int indexes")
if i < 0 or i >= len(self):
raise IndexError
def __getitem__(self, i):
if i == -1:
return self.null_item
self._check_index(i)
if i >= self._lgt:
data = self._extra[i - self._lgt]
else:
index = self._calculate_index(i)
data = self._data[index : index + self.index_size]
r = self.index_format.unpack(data)
if self._lgt and i == 0:
r = (offset_type(0, gettype(r[0])),) + r[1:]
return r
class IndexObject(BaseIndexObject):
def __init__(self, data):
assert len(data) % self.index_size == 0
self._data = data
self._lgt = len(data) // self.index_size
self._extra = []
def _calculate_index(self, i):
return i * self.index_size
def __delitem__(self, i):
if not isinstance(i, slice) or not i.stop == -1 or i.step is not None:
raise ValueError(b"deleting slices only supports a:-1 with step 1")
i = i.start
self._check_index(i)
self._stripnodes(i)
if i < self._lgt:
self._data = self._data[: i * self.index_size]
self._lgt = i
self._extra = []
else:
self._extra = self._extra[: i - self._lgt]
class PersistentNodeMapIndexObject(IndexObject):
"""a Debug oriented class to test persistent nodemap
We need a simple python object to test API and higher level behavior. See
the Rust implementation for more serious usage. This should be used only
through the dedicated `devel.persistent-nodemap` config.
"""
def nodemap_data_all(self):
"""Return bytes containing a full serialization of a nodemap
The nodemap should be valid for the full set of revisions in the
index."""
return nodemaputil.persistent_data(self)
def nodemap_data_incremental(self):
"""Return bytes containing a incremental update to persistent nodemap
This containst the data for an append-only update of the data provided
in the last call to `update_nodemap_data`.
"""
if self._nm_root is None:
return None
docket = self._nm_docket
changed, data = nodemaputil.update_persistent_data(
self, self._nm_root, self._nm_max_idx, self._nm_docket.tip_rev
)
self._nm_root = self._nm_max_idx = self._nm_docket = None
return docket, changed, data
def update_nodemap_data(self, docket, nm_data):
"""provide full block of persisted binary data for a nodemap
The data are expected to come from disk. See `nodemap_data_all` for a
produceur of such data."""
if nm_data is not None:
self._nm_root, self._nm_max_idx = nodemaputil.parse_data(nm_data)
if self._nm_root:
self._nm_docket = docket
else:
self._nm_root = self._nm_max_idx = self._nm_docket = None
class InlinedIndexObject(BaseIndexObject):
def __init__(self, data, inline=0):
self._data = data
self._lgt = self._inline_scan(None)
self._inline_scan(self._lgt)
self._extra = []
def _inline_scan(self, lgt):
off = 0
if lgt is not None:
self._offsets = [0] * lgt
count = 0
while off <= len(self._data) - self.index_size:
start = off + self.big_int_size
(s,) = struct.unpack(
b'>i',
self._data[start : start + self.int_size],
)
if lgt is not None:
self._offsets[count] = off
count += 1
off += self.index_size + s
if off != len(self._data):
raise ValueError(b"corrupted data")
return count
def __delitem__(self, i):
if not isinstance(i, slice) or not i.stop == -1 or i.step is not None:
raise ValueError(b"deleting slices only supports a:-1 with step 1")
i = i.start
self._check_index(i)
self._stripnodes(i)
if i < self._lgt:
self._offsets = self._offsets[:i]
self._lgt = i
self._extra = []
else:
self._extra = self._extra[: i - self._lgt]
def _calculate_index(self, i):
return self._offsets[i]
def parse_index2(data, inline, revlogv2=False):
if not inline:
cls = IndexObject2 if revlogv2 else IndexObject
return cls(data), None
cls = InlinedIndexObject2 if revlogv2 else InlinedIndexObject
return cls(data, inline), (0, data)
class Index2Mixin(object):
index_format = revlog_constants.INDEX_ENTRY_V2
index_size = revlog_constants.INDEX_ENTRY_V2.size
null_item = (0, 0, 0, -1, -1, -1, -1, nullid, 0, 0)
def replace_sidedata_info(self, i, sidedata_offset, sidedata_length):
"""
Replace an existing index entry's sidedata offset and length with new
ones.
This cannot be used outside of the context of sidedata rewriting,
inside the transaction that creates the revision `i`.
"""
if i < 0:
raise KeyError
self._check_index(i)
sidedata_format = b">Qi"
packed_size = struct.calcsize(sidedata_format)
if i >= self._lgt:
packed = _pack(sidedata_format, sidedata_offset, sidedata_length)
old = self._extra[i - self._lgt]
new = old[:64] + packed + old[64 + packed_size :]
self._extra[i - self._lgt] = new
else:
msg = b"cannot rewrite entries outside of this transaction"
raise KeyError(msg)
class IndexObject2(Index2Mixin, IndexObject):
pass
class InlinedIndexObject2(Index2Mixin, InlinedIndexObject):
def _inline_scan(self, lgt):
sidedata_length_pos = 72
off = 0
if lgt is not None:
self._offsets = [0] * lgt
count = 0
while off <= len(self._data) - self.index_size:
start = off + self.big_int_size
(data_size,) = struct.unpack(
b'>i',
self._data[start : start + self.int_size],
)
start = off + sidedata_length_pos
(side_data_size,) = struct.unpack(
b'>i', self._data[start : start + self.int_size]
)
if lgt is not None:
self._offsets[count] = off
count += 1
off += self.index_size + data_size + side_data_size
if off != len(self._data):
raise ValueError(b"corrupted data")
return count
def parse_index_devel_nodemap(data, inline):
"""like parse_index2, but alway return a PersistentNodeMapIndexObject"""
return PersistentNodeMapIndexObject(data), None
def parse_dirstate(dmap, copymap, st):
parents = [st[:20], st[20:40]]
# dereference fields so they will be local in loop
format = b">cllll"
e_size = struct.calcsize(format)
pos1 = 40
l = len(st)
# the inner loop
while pos1 < l:
pos2 = pos1 + e_size
e = _unpack(b">cllll", st[pos1:pos2]) # a literal here is faster
pos1 = pos2 + e[4]
f = st[pos2:pos1]
if b'\0' in f:
f, c = f.split(b'\0')
copymap[f] = c
dmap[f] = e[:4]
return parents
def pack_dirstate(dmap, copymap, pl, now):
now = int(now)
cs = stringio()
write = cs.write
write(b"".join(pl))
for f, e in pycompat.iteritems(dmap):
if e[0] == b'n' and e[3] == now:
# The file was last modified "simultaneously" with the current
# write to dirstate (i.e. within the same second for file-
# systems with a granularity of 1 sec). This commonly happens
# for at least a couple of files on 'update'.
# The user could change the file without changing its size
# within the same second. Invalidate the file's mtime in
# dirstate, forcing future 'status' calls to compare the
# contents of the file if the size is the same. This prevents
# mistakenly treating such files as clean.
e = dirstatetuple(e[0], e[1], e[2], -1)
dmap[f] = e
if f in copymap:
f = b"%s\0%s" % (f, copymap[f])
e = _pack(b">cllll", e[0], e[1], e[2], e[3], len(f))
write(e)
write(f)
return cs.getvalue()