On Tue, Jun 21, 2011 at 07:18:08PM +0200, Enrico Zini wrote: > The locking strategy of u-a-x-i is would be attach to an already running > indexer instead of raising an error, so before applying this patch I'd > like to see if there isn't something else to fix instead.
Indeed there's a bug somewhere: the server isn't, for all practical
purposes, locking the lockfile.
The problem is here:
def lock(self):
[...]
lockfd = os.open(axi.XAPIANDBLOCK, os.O_RDWR | os.O_CREAT)
lockpyfd = os.fdopen(lockfd)
try:
fcntl.lockf(lockpyfd, fcntl.LOCK_EX | fcntl.LOCK_NB)
# Wrap the current progress with the server sender
self.progress = ServerProgress(self.progress)
return True
except IOError, e:
if e.errno == errno.EACCES or e.errno == errno.EAGAIN:
return False
else:
raise
Since no reference is kept to lockfd, the lock expires when the garbage
collector decides it.
I'll now fix that instead of applying your patch. I hope you don't mind
:)
Ciao,
Enrico
--
GPG key: 4096R/E7AD5568 2009-05-08 Enrico Zini <enr...@enricozini.org>
signature.asc
Description: Digital signature
