[Qemu-devel] QEMU summit 2013 minutes

[Qemu-devel] QEMU summit 2013 minutes

[Luiz Capitulino]

Hi,

This year's QEMU summit was held on October 21st, during KVM Forum in
Edinburgh, Scotland.

I volunteered to take the meeting's minutes, which you'll find below. But
it was harder than I expected to do it, so please allow for corrections
from other people who joined the meeting.


o Attendees:

 - Anthony Liguori
 - Kevin Wolf
 - Luiz Capitulino
 - Michael S. Tsirkin
 - Michael Roth
 - Peter Maydell
 - Gerd Hoffmann
 - Max Filippov
 - Alex Graf
 - Stefano Stabellini
 - Andreas Färber
 - Paolo Bonzini
 - Stefan Hajnoczi
 - Juan Quintela


o Software Freedom Conservancy
 - QEMU is considering to become a project member
 - To help with legal questions
 - Financial account (to accept donations and payments for GSoC)
 - Interest in funding OPW interns in the future
 - Need to document a project organization structure
 - Have to appoint a board (4 - 8 peoples)
  * Anthony needs to clarify what obligations come with board participation
    - In order for individuals to obtain approval from their employers
  * Conservancy does not "own" QEMU in the sense that it owns any
    copyright to the project.  It effectively manages the project
 - Are there alternatives?
   * Anthony looked at it: it's a different kind of non-profit
 - Action items: Anthony will fill the application and try to get
   it submitted in November

o State of maintainership
 - QEMU keeps growing
   * adding more people and existing people are writting more code
   * How can we fix patch merging, and patch quality?
 - Some areas are not cared, because they are not people's personal interest
 - Michael T.: Discuss on the list & fix the MAINTAINERS file
 - Some people with commit access disappeared from the project
 - Anthony was busy moving
 - Anybody can review or setup a tree
 - Paolo: people should send their own pull requests?
   * Get reviewed-bys and send pull requests
   * Random people shouldn't send pull requests, better to get Reviewed-bys
 - Anthony: Require patches to have at least one Reviewed-by
   * Should submaintainers add Reviewed-by in addition to Signed-off-by?
     - No native git-am / git-commit support
   * Andreas: When people step up as maintainer of not actively
     maintained area, unlikely to get Reviewed-bys from others!
     (e.g., CPU refactorings by Andreas with him as maintainer)
   * Andreas: Sometimes just informal "looks OK" rather than Reviewed-by
     - Shall be recorded as Acked-by, not as Reviewed-by
 - Signed pull requests
   * Will start with 1.7
   * Required by 2.1?

o Testing
 - Does buildbot still work?
   * It's in a bad state right now, just doesn't work
   * Our infrastructure doesn't work
   * Stefan: there's another better tool for it: Travis
     - It has some limitations (?)
   * Stefan: we need someone to pursue and maintain this
   * Should have: distributed, people could contribute their hardware,
     have the results on qemu.org
   * Anthony will push on maintainers to have automated testing
     - May refuse pull requests w/o automated tests
 - Andreas: Basic qtest coverage of machines prepared
   * Discuss remaining problems in Hackathon?
   * Call for maintainers to add at least trivial qtests for existing
     non-default devices to assure they don't break

o Google Summer of Code
 - Did it work? Can we improve it?
 - Stefan: I don't get feedback from the community
   * If projects are not merged, qemu doesn't benefit
   * Communication with some students had problems (need to improve)
   * Two students became qemu developers
   * Students who failed was because they didn't have the requeriments
   * Juan: information on wikipage has to be clearer on the project's
     requeriments
   * Peter: warn students they may fail before they do
   * Alex: get patches posted upstream sooner
   * Paolo: remember to be tactful when failing students
   * This year we donated mentor's stipends to the Tor project

o The tone on the ML, is it getting bad?
 - Paolo thinks it's the increased amount of emails
 - Some projects have formal code of conduct
   * This can be overkill for us
 - Stefano: we need good examples
   * need to have everyone doing it
 - How can we improve it?
 - Need someone to mediate when there's a problem

o Should we have a merge window dev style?
 - Alex: doesn't change anything
 - Paolo: benefit: can free Anthony during stablization period
 - Juan: external people can help testing
 - Merging several trees at the same time can generate problems
 - Anthony: we'll extend release cycle, people hate it
 - We may get there eventually
 - We have cleanup to do in our process before making changes
   * Increase hard freeze period (3 or 4 weeks?)

o Andreas: who should be the responsbility to propose to -stable? Maintainers
 or author?
 - General opnion: both
 - Alex: Greg has scripts to collect patches for -stable
 - Stefano: we should clear policy on older version releases
   < FIXME: Anthony explained it, but Luiz couldn't pay attention & write
     at the same time. Ask Anthony to write on the wiki >
 - People can vonlunteer to maintan older trees if they want
   * Only makes sense if you have an investiment (eg a distro)

o Andreas: Inconsistent prefixes in commit messages lately
  (e.g., target-i386 vs. x86, s390/... vs. s390x, hw/... or not)
 - Consistent prefix scheme desired for working with commit history
   (as opposed to submaintainers' individual mail filtering processes)
   * Can be sanitized by submaintainers as desired (no consensus)