phases: avoid N² behavior in `advanceboundary`
We allowed duplicated entries in the deque, which each entry could potentially
insert all its ancestors. So advancing boundary for the full repository would
mean each revision would walk all its ancestors, resulting in O(N²) iteration.
For repository of any decent size, N² is quickly insane.
We introduce a simple set to avoid this and get back to reasonable performance.
#!/usr/bin/env python
"""
Tests the behavior of filelog w.r.t. data starting with '\1\n'
"""
from mercurial.node import hex
from mercurial import (
hg,
ui as uimod,
)
myui = uimod.ui.load()
repo = hg.repository(myui, path=b'.', create=True)
fl = repo.file(b'foobar')
def addrev(text, renamed=False):
if renamed:
# data doesn't matter. Just make sure filelog.renamed() returns True
meta = {b'copyrev': hex(repo.nullid), b'copy': b'bar'}
else:
meta = {}
lock = t = None
try:
lock = repo.lock()
t = repo.transaction(b'commit')
node = fl.add(text, meta, t, 0, repo.nullid, repo.nullid)
return node
finally:
if t:
t.close()
if lock:
lock.release()
def error(text):
print('ERROR: ' + text)
textwith = b'\1\nfoo'
without = b'foo'
node = addrev(textwith)
if not textwith == fl.read(node):
error('filelog.read for data starting with \\1\\n')
if fl.cmp(node, textwith) or not fl.cmp(node, without):
error('filelog.cmp for data starting with \\1\\n')
if fl.size(0) != len(textwith):
error(
'FIXME: This is a known failure of filelog.size for data starting '
'with \\1\\n'
)
node = addrev(textwith, renamed=True)
if not textwith == fl.read(node):
error('filelog.read for a renaming + data starting with \\1\\n')
if fl.cmp(node, textwith) or not fl.cmp(node, without):
error('filelog.cmp for a renaming + data starting with \\1\\n')
if fl.size(1) != len(textwith):
error('filelog.size for a renaming + data starting with \\1\\n')
print('OK.')