On Sun, Feb 19, 2017 at 10:24 PM, Mark Hammond <[email protected]> wrote:
> Work has been proceeding well on the desktop bookmark repair work (bug > 1317223 <https://bugzilla.mozilla.org/show_bug.cgi?id=1317223>). The > current status is: > > - All current work is on the elm twig > <https://treeherder.mozilla.org/#/jobs?repo=elm>, which is > continuously running the full test suite. This is regularly being updated > against mozilla-central, so no rebasing surprises are expected. > - Almost all patches have been fully reviewed, with a couple of > exceptions: > - rnewman has a review request on "part 1", which I expect he will > complete soon, and while he might raise some issues I've no reason to > believe they will be difficult to resolve. > - Thom is still working on a couple of patches we've identified we > need - one to prevent multiple requests being started at the same time > from > different devices, and another to ensure we don't repair while in an > inconsistent state. I expect these to be done early this week. > - Kit is working on integration tests, which are proving a little > tricker. If this looks like not being complete this week we may consider > landing without these tests, but treat the tests as a high-priority > after > landing. > - The code is setup such that the code which *initiates *a repair will > not ride the trains past Aurora (although the code that *responds* > will ride the trains) > - The repair process writes detailed event telemetry and is controlled > via a single preference, so we can easily disable the entire repair process > is necessary, and should be able to get good insights into how successful > the repair is. > > Our plan is to land this in the next week or so, at which time we will do > the following: > > > - Update the documentation for the repair process as a guide for the > iOS implementation. > - Arrange to see telemetry for repairs that happen and monitor the > success. > - Work on identified followups > - Tweak the process based on telemetry results > - profit? > > Please let me know if there are any questions about this. > I just wanted to say that I did have lots of questions, but most of them were answered by the excellent technical planning document you posted to the ticket. Great work! My outstanding question: if Sync were a different system, we might handle this scenario by migrating the client collections forward to a new collection version. That functionality has been punted down the road so many times it's black and blue, but did you write the rationale for why it's not feasible? It's an argument that I mostly understand but have never seen written down. Best, Nick
_______________________________________________ Sync-dev mailing list [email protected] https://mail.mozilla.org/listinfo/sync-dev
