[Chandler-dev] [Sum] Dec 4 - 10

[Katie Capps Parlante]

Build, Release and QA
---------------------
*SVK*
Grant sent out notes on how he's using SVK. Vinu described his scenario 
using SVK, and pointed out that people on #svk on freenode are helpful.
http://wiki.osafoundation.org/bin/view/Journal/GrantBaillieSvkNotes
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007358.html

*Alpha5 Checkpoint*
Bear spins the Dec 4th checkpoint:
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007355.html
Dan tests:
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007371.html

*Analysis of 0.7alpha4 feedback data*
Heikki summarized the feedback data from 0.7alpha4 submissions. Time 
between failures was lower compared to previous analysis -- Heikki 
theorized this was from people outside the office who don't know how to 
avoid various bugs/problems. He also noted that Linux is well 
represented in the logs, as well as Windows 2000. He listed out some of 
the errors, looking for feedback if any were known/expected. Heikki 
logged them as bugs, noting that this may end up being a regular QA task 
in the future.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007369.html 


*Python 2.5*
Bear gave a heads up that the Python 2.5 upgrade was about to land in 
the trunk. Andi did the work in a branch. Bear listed out packages 
requiring changes.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007378.html

Heikki did performance tests comparing the Python 2.5 branch with the 
trunk, using Ubuntu Linux. He didn't notice a meaningful difference. 
More details here:
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007356.html

Grant pointed out that single measurements aren't particularly 
trustworthy, as he often sees 5-10% variations in performance test 
results, probably due to some random OS task.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007377.html

The upgrade landed, requiring 'make realclean' to remove old .py[co] 
files. For those building external, 'make realcleanworld'. John found 
and logged some bugs with realclean.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007387.html

*Tinderbox logs*
Heikki noted that people don't like the popup dialog that interferes 
with checking build logs from Tinderbox. Heikki finds one bit if info 
useful from that dialog: the time it took for the cycle. Heikki gave 
instructions on how to work around the dialog, noting browser/platform 
differences. He asked about better solutions.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007381.html

Grant pointed out that the other information in the dialog is 
useless/redundant. He suggested including the cycle time as a regular 
html element. Grant finds the current behavior annoying and would be 
willing to help fix it.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007382.html

*QA IRC sessions*
Aparna proposed taking over the 11 am IRC slot for QA test sessions, 
starting in January. Plans include collaborative testing for new 
features in Cosmo and Chandler, bug verification, regression testing, 
test framework discussions, etc. The session on Jan 3 will focus on 
Cosmo, in advance of the 0.6 release.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007404.html

*Instruction manual for dogfooders*
Aparna wrote up instructions for dogfooders, and asked for feedback. 
Davor asked about privacy issues, as the instructions suggest posting an 
exported calendar.
http://wiki.osafoundation.org/bin/view/Projects/InstructionsForDogfooders
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007408.html

Mikeal suggested adding a feature to the feedback feature, allowing 
people to optionally send along data sets (exported ics file, 
repository). The data would then be accessible to bugfixers internally, 
but not public in Bugzilla.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007409.html

Dev
---
*Edit/Update*
We had a couple of threads on the edit/update topic, as well as a 
pivotal irc conversation. I'll summarize these in a separate message, 
but threads start here:
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007361.html
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007392.html

*Network operations for subscribe*
Morgen wrote up the algorithm for subscribe in pseudocode. He wanted 
people to take a look at the network operations in particular. He noted 
one problem: Cosmo responds to a GET request on a CalDAV calendar with 
an iCalendar file of all contained events. Given Morgen's planned 
approach, this causes a problem for Chandler being able to subscribe to 
a Cosmo calendar in CalDAV mode (in particular the ability to detect the 
difference between CalDAV calendars and monolithic iCalendar files). 
Morgen didn't include logic for detecting write permission. He asked if 
we could do ticket discovery to find permissions. His current trick is 
to MKCOL a temporary subcollection to verify write permission.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007363.html

Brian Moseley replied that Cosmo will have separate urls for webcal and 
dav. Webcal GET will return a monolithic iCalendar resource, and DAV GET 
will return an html directory listing. He explained that CalDAV has a 
mechanism for determining if a resource is a calendar collection: 
PROPFIND the url and look at the DAV:resourcetype property. He also 
explained how to use PROPFIND to discover the ticket permissions 
(ticket:ticketdiscovery property). He also noted that Morgen's MKCOL 
trick isn't guaranteed to work -- calendar collections are not 
guaranteed to support subcollections.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007366.html

Grant explained that a GET on Apple's server would also do a large 
download and miss out on optimal CalDAV behavior. Grant suggested that 
Chandler do an initial PROPFIND, asking for ACL/privileges, 
getcontenttype, resourcetype. If that fails, revert to webcal mode and 
GET the url. Grant noted that "text/calendar" is the correct contenttype 
for iCalendar.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007374.html

Grant and Jeffrey explained that zanshin should support the 
ACL/privileges PROPFIND as of Bug 7297. Zanshin's getPrivileges method 
should check the ticket:ticketdiscovery property.
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007372.html

Grant asked how we know if a DAV url supports eim + xml (eimml). Brian 
replied that we won't be implementing eimml over DAV (morse code is the 
  current name for the protocol).
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007375.html

*Mail check in*
Brian Kirsch checked in a major mail code enhancement, including schema 
changes as part of r12584.
- fromMe/toMe booleans on MailStamp, dateSent is now shared
- improved status bar messages when downloading/sending
- modal dialog w/progress meter for mail account testing
- In/Out collection fixes (Bug 5988, Bug 6857, Bug 6849). In/Out 
collections now calculated based on whether email is from me or to me.
- meAddressCollection contains all current and past "me" addresses, and 
is used for an expanded notion of "me". (For debugging, the list is 
accessible from a test menu item.)
- isOutbound flag defaults to false (not all newly created mail is outbound)
- new unit test (TestInOutCollections)
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007397.html

Meetings, Summaries, Announcements
----------------------------------
Aparna's QA summary Nov 27 - Dec 1:
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007357.html
and Dec 4 - 8:
http://lists.osafoundation.org/pipermail/chandler-dev/2006-December/007405.html

No apps meeting, status:
http://wiki.osafoundation.org/bin/view/Journal/AppsMeeting20061207