keepalive: ensure `close_all()` actually closes all cached connections
While debugging why LFS blob downloads are getting corrupted with workers, I
noticed that prior to spinning up the workers, the ConnectionManager has 2
connections to the server and calling `KeepAliveHandler.close_all()` left one
behind. The reason is the value component of `self._cm.get_all().items()` is a
list, and `self._cm.remove()` modifies said list while the caller is iterating
over it. Now `get_all()` is a deep copy of both the dict and lists in all
cases.
# Disable the $CAP wire protocol capability.
if test -z "$CAP"
then
echo "CAP environment variable not set."
fi
cat > notcapable-$CAP.py << EOF
from mercurial import extensions, localrepo
from mercurial.interfaces import repository
def extsetup(ui):
extensions.wrapfunction(repository.peer, 'capable', wrapcapable)
extensions.wrapfunction(localrepo.localrepository, 'peer', wrappeer)
def wrapcapable(orig, self, name, *args, **kwargs):
if name in b'$CAP'.split(b' '):
return False
return orig(self, name, *args, **kwargs)
def wrappeer(orig, self):
# Since we're disabling some newer features, we need to make sure local
# repos add in the legacy features again.
return localrepo.locallegacypeer(self)
EOF
echo '[extensions]' >> $HGRCPATH
echo "notcapable-$CAP = `pwd`/notcapable-$CAP.py" >> $HGRCPATH