Merge lp:~mandel/ubuntuone-client/stale-broker into lp:ubuntuone-client
Proposed by
Manuel de la Peña
Status: | Merged | ||||
---|---|---|---|---|---|
Approved by: | Mike McCracken | ||||
Approved revision: | 1297 | ||||
Merged at revision: | 1299 | ||||
Proposed branch: | lp:~mandel/ubuntuone-client/stale-broker | ||||
Merge into: | lp:ubuntuone-client | ||||
Diff against target: |
200 lines (+113/-3) 3 files modified
tests/platform/tools/test_tools.py (+64/-0) ubuntuone/platform/ipc/ipc_client.py (+12/-0) ubuntuone/platform/tools/perspective_broker.py (+37/-3) |
||||
To merge this branch: | bzr merge lp:~mandel/ubuntuone-client/stale-broker | ||||
Related bugs: |
|
Reviewer | Review Type | Date Requested | Status |
---|---|---|---|
Mike McCracken (community) | Approve | ||
Roberto Alsina (community) | Approve | ||
Review via email: mp+121456@code.launchpad.net |
Commit message
- Fix stale broker problems by reconnecting to the server and requesting new valid instances for the reference objects in the client side (LP: #1040915).
Description of the change
- Fix stale broker problems by reconnecting to the server and requesting new valid instances for the reference objects in the client side (LP: #1040915).
To post a comment you must log in.
As discussed on IRC, I'd like to understand a bit more about why we're disconnecting and getting invalid refs before we just auto-reconnect on failures.
The bug seems to come up when switching between tabs, and we might (didn't confirm) have one PB connection per tab to the syncdaemon - is something locally getting GC'd when a tab isn't visible?
If so, why does that cause a remote ref to be invalid? I haven't seen any evidence that syncdaemon itself is dying.