Once again, two weeks combined into one summary...
Build, Release and QA
---------------------
*Checkpoints*
Bear spins (Jan 8th checkpoint):
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007491.html
Dan tests:
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007494.html
Bear spins again (Jan 16):
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007504.html
Dan tests:
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007512.html
*wxPython*
Robin updated Chandler's wxPython, merging wxPython 2.8.10 with Chandler
customizations and adding some bug fixes.
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007500.html
*Test case specs*
Dan pointed to test case specs on the wiki for email, recurrence, and
triage.
http://svn.osafoundation.org/docs/trunk/testspecs/rel0_7/Email-0.7-testspec.html
http://svn.osafoundation.org/docs/trunk/testspecs/rel0_7/Recurrence-0.7-testspec.html
http://svn.osafoundation.org/docs/trunk/testspecs/rel0_7/Triage-0.7-testspec.html
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007507.html
*Test server*
Jared set up "qasharing.osafoundation.org" as a Cosmo test server,
dedicated to being a target for Chandler sharing tests. Previously we'd
been using the production server -- not a good long term solution.
"qacosmo" is updated regularly for Cosmo testing purposes, "qasharing"
will be more stable. The server is outside the office, to better
simulate real world sharing conditions. Jared had a few requests:
- we find a maintainer
- we upgrade to pre-rc or rc of Cosmo 0.6 in the next 5/10 days
- move tinderbox tests away from osaf.us
- move to qasharing as permanent home for sharing tests
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007520.html
Philippe suggested Build/Release should be the maintainer, and the
decision to move to Cosmo 0.6 should be made at the biweekly Chandler
eng meeting.
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007522.html
*pystack*
Grant sent a patch to gdbinit to get pystack working against Chandler's
python.
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007519.html
Dev
---
*Alarm Domain Model*
Katie asked if we could adjust the domain model so that alarms align
better with iCalendar. In particular, a text description of the alarm is
needed (usually a combination of summary + location). Jeffrey agreed we
should flesh this out. (Tracking as Bug 7915).
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007495.html
*Item reference refactor, pruning Items*
Andi replaced 'SingleRef' with 'ItemRef', a smarter implementation.
ItemRef wraps a UUID, a repository view, and a weak reference to an
Item. As Items are cached, dereferencing them is faster. These Items are
not ref-counted, so ItemRefs don't prevent an Item from being 'pruned'
(unloaded from memory). If an Item does get unloaded and is then
re-referenced, the ItemRef will reload it. Andi experimented with
--prune, discovering that the main UI view likes to have ~1500 items
loaded. He explained that much smaller prune sizes make sense for other
background views (email, sharing). Andi observed that Python doesn't
return memory to the OS, but does reuse memory freed by puning Item
instances from a view. Andi asked that devs pay attention to choosing a
good prune size for each view. More detail:
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007497.html
*MVCC performance*
Heikki tried out --mvcc on tinderbox perf tests. The only tests
registering significant changes were subscribing to a collection (15%
faster), and publishing a collection (500% slower). Andi pointed out
that we'd only expect an improvement in cases with concurrent
transactions, so no-news-is-good-news for the bulk of the tests.
(Conversation about the publishing case continues in Bug 7757)
https://bugzilla.osafoundation.org/show_bug.cgi?id=7757
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007506.html
*bye bye redirectTo, bye bye subject?*
Andi removed support for 'redirectTo', changing remaining uses to use
property() or schema.Calculated(). If get/setAttributeValue() is called
on one of these altered attributes, you'll see a NoSuchAttributeError.
(Andi recommends get/setattr for performance reasons in any case).
Shared emails w/out other stamps may lose their 'subject'. 'subject' is
now a schema.Calculated() based on ContentItem.displayName().
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007508.html
Bryan Stearns suggested that we get rid of the notion of 'subject'
internally, treating 'displayName' as the subject (labeling it
appropriately). Brian Kirsch agreed with Bryan Stearns suggestion, but
noted he'd like to save that change until after Preview, along with
other MailStamp schema refactorings.
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007513.html
*Edit/Update*
Philippe is trying to coordinate edit/update work, as it involves
several different people. He summarized a few related meetings and tasks
on the wiki page. Philippe reminded everyone to focus on finding a
workable solution for the common cases. (Note, Edit/Update discussions
have spilled into related topics: conflict management and
recurrence/sharing).
http://wiki.osafoundation.org/Projects/EditUpdateTracking
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007517.html
*Empty values*
After noticing that the 'Welcome Note' had 49 values, Andi did some
experiments to look at ContentItems' empty values that get set up by
initialValue(). In particular, a small number of attributes set up a
large number of empty dicts. For values that aren't references (in
particular booleans), Andi recommended using defaultValue. Grant agreed
with the suggestion, noting some risk given making the changes will
alter the result of a call to hasLocalAttributeValue().
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007518.html
Grant explained approaches used for reflists: (1) defaultValue=None (2)
initialValue=[]. The first involves extra boilerplate code (checking for
None, initializing). The second produces simpler code, but results in
extra attribute assignments. This happens more frequently than it used
to due to the way Stamps are implemented, but there is a bug to fix that
particular case. Grant proposed trying to get the best of both worlds:
the behavior of defaultValue=[] until one tries to modify the collection.
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007521.html
Andi proposed following up on Grant's suggestion, implementing a "Nil"
value that can be iterated and that knows what to do when an attribute
attempts to append to it. Andi noted that preventing Stamp subclasses
from setting initialValues until the Stamp is added would go a long way
to addressing the issue.
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007523.html
Meetings, Announcements, Summaries
----------------------------------
*QA summary Jan 2-5*
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007493.html
*IRC test sessions Jan 10 (Chandler) and Jan 17 (Cosmo)*
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007498.html
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007509.html
*Apps team meeting, status*
http://wiki.osafoundation.org/bin/view/Journal/AppsMeeting20070111
http://wiki.osafoundation.org/Journal/AppsMeeting20070118
*New users list*
For users of Chandler and Cosmo:
http://lists.osafoundation.org/pipermail/chandler-dev/2007-January/007515.html
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Open Source Applications Foundation "chandler-dev" mailing list
http://lists.osafoundation.org/mailman/listinfo/chandler-dev
