[Chandler-dev] [Sum] Nov 6 - 12
Katie Capps Parlante
capps at osafoundation.org
Fri Nov 17 14:50:20 PST 2006
Build, Release, QA
Bear spun a checkpoints on Nov 6, 9, and 10. The checkpoint on the 10th
broke sharing compatibility (see below):
Pieter proposed that we highlight alpha4 instead of 0.6 on the website.
*Road to Alpha4*
Philippe gave some background on why we've been slipping alpha4, and the
path to get to a release candidate. Philippe explained that our
objective with alpha4 is to have a stable release for dogfooders, and
that we have been holding the release until we meet these criteria:
- no alpha4 blocking bugs
- 2 consecutive days without new blocking bugs reported by dogfooders
Blocking bug criteria?
- reproducible crashers
- data loss or loss of major functionality
Please do all dogfooding on alpha4 until we've released!
An additional wrinkle, we decided to go ahead and qualify the release
against osaf.us (Cosmo 0.5), so this added a bit of time to the release
as well. Testing against the public server isn't really the best plan
for the long term, but the server isn't really public yet and the extra
mileage/testing is useful. Eventually we'll need a separate test server
(cosmo-test). More details from a recent QA/Build/Release meeting:
*Chandler update breaks compatibility*
An update on the branch and the trunk broke compatibility. Dogfooders
will need to create new collections on osaf.us. To do so, people will
need to back up data to *.ics files (using an older Chandler) and
reimport them (using the new updated Chandler). This process causes some
loss of data, though it can be mitigated by stamping items as Events.
Philippe gave detailed instructions, and included the tickets to the new
*Calendar out of wack*
Morgen explained that the office calendar got out of wack. Due to a
sequence of events with changes in how UUIDs are generated/handled when
calendars are imported from *.ics files, and the migration steps taken
moving from cosmo-demo to osaf.us, we ended up with parallel sets of the
same items with different UUIDs. Morgen had some fixes to prevent the
situation from happening again:
- detect importing the same events with a different UUID
- bump the xml version number so that old versions don't cause problems
- deterministically generate UUIDs from non-UUID-like iCal UIDs using an
md5 hash (e.g. data imported from Outlook). The upshot is that fixing
the problem necessitated losing backward compatibility w/existing
shares, new shares needed to be created. Morgen's explanation avec gory
*SVN instead of builds server for tarballs*
Heikki suggested we use svn for source and binary tarballs, instead of
placing them on the builds server. Andi was unconvinced that the way we
do it now is broken.
Bryan, Jeffrey and Grant met up to talk about dashboard work in alpha5.
Jeffrey will focus on occurrences-as-modifications in dashboard
collections. Grant will work on per-attribute-modifications ("modified"
occurrences will inherit non-modified attributes from their masters --
Bug 6700). Grant will also take on the update workflow architecture and
*Localizable sample text and displayName*
Reid has a task to implement localizable sample text in text fields (Bug
7292). He asked about 'displayName' going away, as a code comment
suggested a previous strategy was to use an item's 'displayName' as the
default sample text. Bryan explained that the repository mechanism to
generically provide names for any repository item is going away, but
'displayName' as an attribute on ContentItem is *not* going away. A
dynamic mechanism for providing localized sample text for the "title"
field is still needed.
Cosmo/Chandler meeting about sharing:
More information about the chandler-dev