##// 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:

r47387:15c2f922 default
r47651:8759e22f default
Show More
dateutil.py
387 lines | 11.6 KiB | text/x-python | PythonLexer
# util.py - Mercurial utility functions relative to dates
#
# Copyright 2018 Boris Feld <boris.feld@octobus.net>
#
# 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, print_function
import calendar
import datetime
import time
from ..i18n import _
from .. import (
encoding,
error,
pycompat,
)
if pycompat.TYPE_CHECKING:
from typing import (
Callable,
Dict,
Iterable,
Optional,
Tuple,
Union,
)
hgdate = Tuple[float, int] # (unixtime, offset)
# used by parsedate
defaultdateformats = (
b'%Y-%m-%dT%H:%M:%S', # the 'real' ISO8601
b'%Y-%m-%dT%H:%M', # without seconds
b'%Y-%m-%dT%H%M%S', # another awful but legal variant without :
b'%Y-%m-%dT%H%M', # without seconds
b'%Y-%m-%d %H:%M:%S', # our common legal variant
b'%Y-%m-%d %H:%M', # without seconds
b'%Y-%m-%d %H%M%S', # without :
b'%Y-%m-%d %H%M', # without seconds
b'%Y-%m-%d %I:%M:%S%p',
b'%Y-%m-%d %H:%M',
b'%Y-%m-%d %I:%M%p',
b'%Y-%m-%d',
b'%m-%d',
b'%m/%d',
b'%m/%d/%y',
b'%m/%d/%Y',
b'%a %b %d %H:%M:%S %Y',
b'%a %b %d %I:%M:%S%p %Y',
b'%a, %d %b %Y %H:%M:%S', # GNU coreutils "/bin/date --rfc-2822"
b'%b %d %H:%M:%S %Y',
b'%b %d %I:%M:%S%p %Y',
b'%b %d %H:%M:%S',
b'%b %d %I:%M:%S%p',
b'%b %d %H:%M',
b'%b %d %I:%M%p',
b'%b %d %Y',
b'%b %d',
b'%H:%M:%S',
b'%I:%M:%S%p',
b'%H:%M',
b'%I:%M%p',
)
extendeddateformats = defaultdateformats + (
b"%Y",
b"%Y-%m",
b"%b",
b"%b %Y",
)
def makedate(timestamp=None):
# type: (Optional[float]) -> hgdate
"""Return a unix timestamp (or the current time) as a (unixtime,
offset) tuple based off the local timezone."""
if timestamp is None:
timestamp = time.time()
if timestamp < 0:
hint = _(b"check your clock")
raise error.InputError(
_(b"negative timestamp: %d") % timestamp, hint=hint
)
delta = datetime.datetime.utcfromtimestamp(
timestamp
) - datetime.datetime.fromtimestamp(timestamp)
tz = delta.days * 86400 + delta.seconds
return timestamp, tz
def datestr(date=None, format=b'%a %b %d %H:%M:%S %Y %1%2'):
# type: (Optional[hgdate], bytes) -> bytes
"""represent a (unixtime, offset) tuple as a localized time.
unixtime is seconds since the epoch, and offset is the time zone's
number of seconds away from UTC.
>>> datestr((0, 0))
'Thu Jan 01 00:00:00 1970 +0000'
>>> datestr((42, 0))
'Thu Jan 01 00:00:42 1970 +0000'
>>> datestr((-42, 0))
'Wed Dec 31 23:59:18 1969 +0000'
>>> datestr((0x7fffffff, 0))
'Tue Jan 19 03:14:07 2038 +0000'
>>> datestr((-0x80000000, 0))
'Fri Dec 13 20:45:52 1901 +0000'
"""
t, tz = date or makedate()
if b"%1" in format or b"%2" in format or b"%z" in format:
sign = (tz > 0) and b"-" or b"+"
minutes = abs(tz) // 60
q, r = divmod(minutes, 60)
format = format.replace(b"%z", b"%1%2")
format = format.replace(b"%1", b"%c%02d" % (sign, q))
format = format.replace(b"%2", b"%02d" % r)
d = t - tz
if d > 0x7FFFFFFF:
d = 0x7FFFFFFF
elif d < -0x80000000:
d = -0x80000000
# Never use time.gmtime() and datetime.datetime.fromtimestamp()
# because they use the gmtime() system call which is buggy on Windows
# for negative values.
t = datetime.datetime(1970, 1, 1) + datetime.timedelta(seconds=d)
s = encoding.strtolocal(t.strftime(encoding.strfromlocal(format)))
return s
def shortdate(date=None):
# type: (Optional[hgdate]) -> bytes
"""turn (timestamp, tzoff) tuple into iso 8631 date."""
return datestr(date, format=b'%Y-%m-%d')
def parsetimezone(s):
# type: (bytes) -> Tuple[Optional[int], bytes]
"""find a trailing timezone, if any, in string, and return a
(offset, remainder) pair"""
s = pycompat.bytestr(s)
if s.endswith(b"GMT") or s.endswith(b"UTC"):
return 0, s[:-3].rstrip()
# Unix-style timezones [+-]hhmm
if len(s) >= 5 and s[-5] in b"+-" and s[-4:].isdigit():
sign = (s[-5] == b"+") and 1 or -1
hours = int(s[-4:-2])
minutes = int(s[-2:])
return -sign * (hours * 60 + minutes) * 60, s[:-5].rstrip()
# ISO8601 trailing Z
if s.endswith(b"Z") and s[-2:-1].isdigit():
return 0, s[:-1]
# ISO8601-style [+-]hh:mm
if (
len(s) >= 6
and s[-6] in b"+-"
and s[-3] == b":"
and s[-5:-3].isdigit()
and s[-2:].isdigit()
):
sign = (s[-6] == b"+") and 1 or -1
hours = int(s[-5:-3])
minutes = int(s[-2:])
return -sign * (hours * 60 + minutes) * 60, s[:-6]
return None, s
def strdate(string, format, defaults=None):
# type: (bytes, bytes, Optional[Dict[bytes, Tuple[bytes, bytes]]]) -> hgdate
"""parse a localized time string and return a (unixtime, offset) tuple.
if the string cannot be parsed, ValueError is raised."""
if defaults is None:
defaults = {}
# NOTE: unixtime = localunixtime + offset
offset, date = parsetimezone(string)
# add missing elements from defaults
usenow = False # default to using biased defaults
for part in (
b"S",
b"M",
b"HI",
b"d",
b"mb",
b"yY",
): # decreasing specificity
part = pycompat.bytestr(part)
found = [True for p in part if (b"%" + p) in format]
if not found:
date += b"@" + defaults[part][usenow]
format += b"@%" + part[0]
else:
# We've found a specific time element, less specific time
# elements are relative to today
usenow = True
timetuple = time.strptime(
encoding.strfromlocal(date), encoding.strfromlocal(format)
)
localunixtime = int(calendar.timegm(timetuple))
if offset is None:
# local timezone
unixtime = int(time.mktime(timetuple))
offset = unixtime - localunixtime
else:
unixtime = localunixtime + offset
return unixtime, offset
def parsedate(date, formats=None, bias=None):
# type: (Union[bytes, hgdate], Optional[Iterable[bytes]], Optional[Dict[bytes, bytes]]) -> hgdate
"""parse a localized date/time and return a (unixtime, offset) tuple.
The date may be a "unixtime offset" string or in one of the specified
formats. If the date already is a (unixtime, offset) tuple, it is returned.
>>> parsedate(b' today ') == parsedate(
... datetime.date.today().strftime('%b %d').encode('ascii'))
True
>>> parsedate(b'yesterday ') == parsedate(
... (datetime.date.today() - datetime.timedelta(days=1)
... ).strftime('%b %d').encode('ascii'))
True
>>> now, tz = makedate()
>>> strnow, strtz = parsedate(b'now')
>>> (strnow - now) < 1
True
>>> tz == strtz
True
>>> parsedate(b'2000 UTC', formats=extendeddateformats)
(946684800, 0)
"""
if bias is None:
bias = {}
if not date:
return 0, 0
if isinstance(date, tuple):
if len(date) == 2:
return date
else:
raise error.ProgrammingError(b"invalid date format")
if not formats:
formats = defaultdateformats
date = date.strip()
if date == b'now' or date == _(b'now'):
return makedate()
if date == b'today' or date == _(b'today'):
date = datetime.date.today().strftime('%b %d')
date = encoding.strtolocal(date)
elif date == b'yesterday' or date == _(b'yesterday'):
date = (datetime.date.today() - datetime.timedelta(days=1)).strftime(
r'%b %d'
)
date = encoding.strtolocal(date)
try:
when, offset = map(int, date.split(b' '))
except ValueError:
# fill out defaults
now = makedate()
defaults = {}
for part in (b"d", b"mb", b"yY", b"HI", b"M", b"S"):
# this piece is for rounding the specific end of unknowns
b = bias.get(part)
if b is None:
if part[0:1] in b"HMS":
b = b"00"
else:
# year, month, and day start from 1
b = b"1"
# this piece is for matching the generic end to today's date
n = datestr(now, b"%" + part[0:1])
defaults[part] = (b, n)
for format in formats:
try:
when, offset = strdate(date, format, defaults)
except (ValueError, OverflowError):
pass
else:
break
else:
raise error.ParseError(
_(b'invalid date: %r') % pycompat.bytestr(date)
)
# validate explicit (probably user-specified) date and
# time zone offset. values must fit in signed 32 bits for
# current 32-bit linux runtimes. timezones go from UTC-12
# to UTC+14
if when < -0x80000000 or when > 0x7FFFFFFF:
raise error.ParseError(_(b'date exceeds 32 bits: %d') % when)
if offset < -50400 or offset > 43200:
raise error.ParseError(_(b'impossible time zone offset: %d') % offset)
return when, offset
def matchdate(date):
# type: (bytes) -> Callable[[float], bool]
"""Return a function that matches a given date match specifier
Formats include:
'{date}' match a given date to the accuracy provided
'<{date}' on or before a given date
'>{date}' on or after a given date
>>> p1 = parsedate(b"10:29:59")
>>> p2 = parsedate(b"10:30:00")
>>> p3 = parsedate(b"10:30:59")
>>> p4 = parsedate(b"10:31:00")
>>> p5 = parsedate(b"Sep 15 10:30:00 1999")
>>> f = matchdate(b"10:30")
>>> f(p1[0])
False
>>> f(p2[0])
True
>>> f(p3[0])
True
>>> f(p4[0])
False
>>> f(p5[0])
False
"""
def lower(date):
# type: (bytes) -> float
d = {b'mb': b"1", b'd': b"1"}
return parsedate(date, extendeddateformats, d)[0]
def upper(date):
# type: (bytes) -> float
d = {b'mb': b"12", b'HI': b"23", b'M': b"59", b'S': b"59"}
for days in (b"31", b"30", b"29"):
try:
d[b"d"] = days
return parsedate(date, extendeddateformats, d)[0]
except error.ParseError:
pass
d[b"d"] = b"28"
return parsedate(date, extendeddateformats, d)[0]
date = date.strip()
if not date:
raise error.InputError(
_(b"dates cannot consist entirely of whitespace")
)
elif date[0:1] == b"<":
if not date[1:]:
raise error.InputError(_(b"invalid day spec, use '<DATE'"))
when = upper(date[1:])
return lambda x: x <= when
elif date[0:1] == b">":
if not date[1:]:
raise error.InputError(_(b"invalid day spec, use '>DATE'"))
when = lower(date[1:])
return lambda x: x >= when
elif date[0:1] == b"-":
try:
days = int(date[1:])
except ValueError:
raise error.InputError(_(b"invalid day spec: %s") % date[1:])
if days < 0:
raise error.InputError(
_(b"%s must be nonnegative (see 'hg help dates')") % date[1:]
)
when = makedate()[0] - days * 3600 * 24
return lambda x: x >= when
elif b" to " in date:
a, b = date.split(b" to ")
start, stop = lower(a), upper(b)
return lambda x: x >= start and x <= stop
else:
start, stop = lower(date), upper(date)
return lambda x: x >= start and x <= stop