[FFe] Qt 5.2

What's blocking this from landing now? This has waited far too long in my opinion. I'm afraid we have so many rules about the development release working perfectly now that it's almost impossible to use for development. This is in Debian Testing already.

You're correct that development release is now kept releasable each day, and each daily Ubuntu Touch image should be flawless (http://ci.ubuntu.com/smokeng/trusty/touch/). The new "CI Train" integration does not allow code merges to trunks unless they don't regress those results.

The effects of the development method may not have been discussed enough after the "CI Airline" vUDS discussion in November, from which the current method has evolved. Everything is now worked in silos until they are perfect, and there are tools to allow that. With Qt it's more manual ensuring but still similar to smaller components. Parallel work until all but polishing problems are fixed.

The current Qt 5.2.x blockers include:
- QA team has started testing 5.2.0, but only this week we've gotten tools to create system images using the PPA to be able to get more automated testing
- No multimedia related apps work yet because of lacking qtmultimedia-touch (Multimedia team is working on this)
- Some critical bugs making Touch images unusable like the inability to turn the screen back on once it has blanked
- The V4 engine is seen a bit risky considering the already identified bugs, so the complete 5.2.1 release seems better all around
- 5.2.1 has not been tested yet by QA team. The packaging syncs I've done with 5.2.1 may cause some rebuild problems since the mkspecs directory changed.
- Debian has part of the 5.2.1 in experimental now, with more coming. I'd like to see all of 5.2.1 syncable from Debian for those modules that can be synced (most of them).

Lastly, the process of uploading and building the 20 Qt modules and rebuilding 80 source packages, taking into account the dependencies and architectures, may take several days in case there are any last minute surprises that weren't noticed even with careful preparation.

All in all, 2.5 weeks from now sounds like something when pieces could fall together. I'm putting 5.2.1 into use right now, which should be the start of the final push.

Starting next cycle, Kubuntu will be a major user of Qt5. Landing the major
Qt5 revision just days before, or maybe even after, feature freeze, is totally
unworkable. I think we're likely not to introduce KDE Frameworks 5 packages
in this development cycle, but to do post-release backports once it's more
mature, so it turns out not to have a major impact for 14.04, but something
better will have to be thought out for "U".

I know this it not a workable solution for U. This is the first ever major upgrade of Qt 5, because the Qt 5.1.1 already didn't happen despite plans. 5.2 on the other hand is too important to not have. The Qt 5.3 timeframe is such that we shouldn't delay moving to that in U, but just bite the bullet as soon as possible (I expect KDE Frameworks 5 will also move to 5.3). 5.3 should also be about "stabilization and performance" according to upstream, which sounds good; the QML engine rewrite in 5.2 was a quite ambitious project.

Hi Timo, which of the blockers are still relevant? When do you expect it to be ready?

Dmitry: The relevant blockers are autopilot test failures and multimedia on Ubuntu Touch. Some manual autopilot testing with phablet-test-run and Nexus 4 can be seen at http://is.gd/JA1jg1. The latest news on multimedia was that they know what needs to be done, but I don't have the latest update from end of last week.

Additionally there is a handful of new build or unit test failures with Qt 5.2.1 visible in the PPA, which look minor but need small fixes.

Fixed blockers include Qt 5.2.1 found to fix the "can't unblank after blanking" bug and probably others that were lurking with 5.2.0. Many bugs like crash on start of Dialer + Address book apps have been fixed. Last week Debian also got up to speed with 5.2.1 so now syncing should be possible to do properly.

People needed for the remaining fixes should be available this week, but I will need some confirmations on above issues (when out of my own hands) before making an updated prediction on the readiness. End of this week should still be possible if the Qt 5.2.1 is validated as good enough to fix remaining Touch problems later, although next week seems more likely.

This week is seeing great progress, and the ETA would be now set to March 6th, if the FFe is accepted.

The following remaining issues have been fixed or about to be fixed:

- qtvideo-node has been fixed, a manual upload to PPA done while merge is pending
- qtmultimedia-opensource-src-touch has a fix too and a manual upload to the PPA - about to be tested
- newly found build problems fixed in eg. telephony-service, qtorganizer5-eds
- regarding mkspecs move, maliit-framework and pyqt5 will pose no problem (at least they can be fixed after Qt has been uploaded, if not updated before for 'just do a rebuild' functionality)
- the Ubuntu plugin for Qt Creator 3.0.1 is ready and published in the PPA for desktop users
- QA team has started testing 5.2.1, and we have also automated autopilot test results (not currently in proper Dashboard style, but visible from Jenkins logs)

The following issues remain:
- various smaller items starting to pop up at https://bugs.launchpad.net/bugs/+bugs?field.tag=qt5.2 like wrong character rendering in certain cases
- a filtered view of the above link http://is.gd/pNalUg - fix all critical and high bugs or demote to be lower importance
- some small build problems may still pop up that are triggered by 5.2.1
- fix remaining autopilot errors as much as possible

The balance will be made between accepting some regressions in case it seems overall there's nothing that can't be fixed.

From a release standpoint, the hurdles put in place of this upload via the CI process exceed anything the release team would normally require for such an FFe. And considering the Ubuntu phone image is currently the predominant consumer of Qt5, there's no reason to be concerned that the CI testing won't adequately cover us. FFe granted.