[ros-dev] March 2012 Meeting Minutes

[Colin Finck]

March 2012 Meeting Minutes

2012-03-29
19:00 UTC
Fezile, #meeting

Proceedings
===========
* Meeting started at 19:00 UTC by Aleksey Bragin.

* Point 1: The state of Trunk
-----------------------------
* Cameron Gutman gave some examples how badly Trunk has regressed since 
0.3.14:
    - Frequent bugchecks and assertions in the console
    - Broken Unicode in the console
    - Several key apps such as Firefox, Ping, Dwnl and Notepad being
      broken or rarely working

   He opened a discussion about improving the quality of Trunk.

* Timo Kreuzer favorized a stricter revert-on-regressions policy. By 
employing the Patchbot buildslave, "evil" changes should be detected in 
advance before committing them. According to Giannis Adamopoulos, the 
Patchbot currently needs 20 minutes in total for building and testing a 
patch.
* Colin Finck asked whether the current regressions were visible in the 
automated regression tests. Aleksey Bragin denied this for most of them 
as these were usability problems not detected by our regression tests.
* Various people suggested frameworks such as AutoIt and AutoHotKey to 
automate application testing as well.
* To also automate real hardware testing, Jan Blomqvist Kinander has set 
up another server. He reported that he could successfully perform a 
remote ReactOS installation on it and just needs some scripts to 
automate the procedures now.
* Aleksey Bragin illustrated his plan to move more towards loosely 
managed development. In particular, he wants to introduce Atlassian 
tools such as FishEye and JIRA to plan activity and actually do stuff.
* Cameron Gutman wanted a firm decision about dealing with regressions 
first. Aleksey Bragin responded that the current rules still apply: A 
commit shall be reverted after one day if it breaks booting or 
installation. Other regressions shall be reported in Bugzilla using the 
REGRESSION prefix.
* Olaf Siejka criticized the way bugs are currently handled. The list at 
http://www.reactos.org/wiki/Bug_Filters#Patch_review_list is getting 
bigger and bigger while developers don't take responsibility for it. He 
added that a decision was taken to release early last month, although no 
progress towards implementing it was visible.
* Alex Ionescu joined the meeting later and reopened the discussion 
about using Atlassian tools. In his opinion, JIRA and FishEye can 
support our development along with Bamboo as a build and continuous 
integration system. Later on, these systems could be set up to only 
accept a commit if it can be successfully built by Bamboo.
    - Olaf Siejka wants to set up a Bamboo test system to evaluate the
      tool.
    - Alex Ionescu will send out an E-Mail as soon as our setup is ready
      for testing.

* Point 2: Plan for the next month
----------------------------------
* The previous discussion about improving development using new tools 
continued here. While Aleksey Bragin outlined that he wants to finish 
the setup in April, Alex Ionescu gave more examples about what needs to 
be changed when moving to JIRA:
    - A dynamic component maintainer list should be added to establish
      a clear chain of responsibility for all parts of ReactOS.
    - User privileges should be changed to let only component maintainers
      classify their bugs.
* Furthermore, the list at 
http://www.reactos.org/wiki/Bug_Filters#Patch_review_list has been 
discussed again. Alex Ionescu and Timo Kreuzer agreed that such general 
lists are pointless and the following approach was suggested instead:
    - Developers owning bugs and being mailed about them
    - Release not being made until all important bugs are fixed
    - Someone to prioritize which bugs are blocking a release
* Alex Ionescu outlined more ideas about improving the release process. 
He suggested to create a roadmap for each release and picking several 
specific bugs as goals for this release. Only fixes for these bugs 
should go into a release, new features could be prepared in branches.
* The branches idea was picked up by Amine Khaldi who questioned that 
SVN was able to properly handle more branches for new features, release 
preparation, etc. due to its limited branching capabilities.
* Alex Ionescu did not agree with this, because people can merge just 
subdirectories of a branch into Trunk or even bundle all changes in 
.patch files and simply apply them.
* Timo Kreuzer called for a feature freeze in Trunk before the 
discussion on point 2 ended, but Aleksey Bragin stated that this would 
prevent further work on fixing Mm.
* Aleksey Bragin finally repeated that he is going to finish the setup 
of the new tools in April and wants to get everybody used to them. The 
idea of having modules owners shall get more developer responsibility 
towards issues and patches.

* Point 3: Next release discussion
----------------------------------
* Aleksey Bragin postponed this discussion to the next meeting as most 
of the topics have already been covered in the discussion about Trunk. 
No objections were raised.

* Point 4: New website status
-----------------------------
* Maciej Bialas reported that he is at the point of applying small fixes 
now as well as importing the user database into Drupal. He stated that 
every user needs to change his password after the migration because of 
different hashing algorithms used in Drupal and RosCMS.
* Pierre Schweitzer noted that we should prevent another mass mailing of 
all registered users due to the problems caused by the last one.
* Ziliang Guo proposed to just create a sticky forum post and write a 
mail to a ReactOS mailing list about the password changes. There were no 
objections.
* The participants agreed to continue discussions about website content 
on the mailing lists.

* Point 5: Miscellaneous
------------------------
* Giannis Adamopoulos picked up the new subsystem layout for Win32k, 
which had already been discussed a year ago and was postponed. He 
outlined it at http://pastie.org/3694259. Aleksey Bragin, James Tabor 
and Timo Kreuzer agreed on it, so it will be committed within the next 4 
days.
* Art Yerkes reported about his work on NewCC. In particular, he is 
currently looking at a bug not existing in a non-NewCC build. He expects 
other issues afterwards, but thinks that the entire work is progressing 
well.

* Meeting closed at 21:55 UTC by Aleksey Bragin.
* Minutes written by Colin Finck.