[Releasing] Meeting minutes: Qt release team meeting 18.08.2015

Heikkinen Jani jani.heikkinen at theqtcompany.com
Wed Aug 19 07:51:31 CEST 2015


Meeting minutes from Qt Release Team meeting 18th August 2015


Qt 5.5 Post Mortem:

- RTA tests to be updated to compare contents of offline installers with previous ones to notice if something is missing from there. AP Jani

- Installer changelog to be added to be able to see what should have changed.  AP Jani

- RTA (smoke set) to be run before manual testing requests to avoid testing all these broken packages. AP Jani

- We need to be stricter on the branch to avoid late changes. AP ALL!


Qt 5.6 Packaging

- In Mac packages will be same than with 5.5

- In widows we will drop MSVC 2010 & MSVC 2012 packages & add MSVC 2015 ones

- In Linux we will drop 32 bit packages


Next meeting Tue 1st September 16:00 CET


br,

Jani


irc log below:


[17:00:50] <jaheikki3> akseli: iieklund: kkoehne: thiago: fkleint: ZapB: tronical: vladimirM: aholza: peter-h: mapaaso: ankokko: fkleint: carewolf: fregl: ablasche: ping
[17:01:09] <thiago> jaheikki3: pong
[17:01:44] <kkoehne> jaheikki3: pong
[17:01:49] <akseli> jaheikki3: pong
[17:01:56] <fkleint> jaheikki3: pong
[17:02:43] <jaheikki3> Time to start Qt release team meeting
[17:02:51] <jaheikki3> On agenda today:
[17:03:09] <jaheikki3> - Qt 5.5 post mortem
[17:03:44] <jaheikki3> - Qt 5.6 packaging changes
[17:03:55] <jaheikki3> any additional item to the agenda?
[17:05:33] <jaheikki3> lets start from qt5.5 post mortem:
[17:05:58] <carewolf> pong
[17:06:12] <jaheikki3> There is few comments in http://wiki.qt.io/Qt_5.5.0_Post_Mortem
[17:06:16] <kkoehne> I think we all agree that 5.5.0 wasn't really a smooth release
[17:06:58] <jaheikki3> kkoehne: At least I agree ;)
[17:07:20] <fkleint> ..but it is already too late to learn from it as 5.6 is basically due tomorrow...
[17:07:23] <fkleint> ;-)
[17:07:38] <carewolf> it is already christmas?
[17:08:18] <kkoehne> A couple of action items for next time: Let's diff the file contents of the offline installers with previous releases
[17:08:45] <thiago> good idea
[17:08:48] <thiago> what prompted this?
[17:08:53] <kkoehne> Also, I think there should be a ChangeLog for the installers itself, so one can easily see what should have changed
[17:08:54] <jaheikki3> fkleint: Well, hoping we could learn at least something ;) One big thing learned is the schedule: With 5.6 it should be more realistic one than earlier at least
[17:09:03] <kkoehne> thiago: E.g. we just missed the .pdb files on Windows packages
[17:10:00] <jaheikki3> kkoehne: yes, we have already discussed it in the team as well. I'll check with RTA team if they are already done a test for it
[17:11:20] <jaheikki3> kkoehne: Installer changelog is good idea as well, I'll take an action point & check it with Iikka
[17:12:32] <thiago> ok
[17:12:35] <jaheikki3> we have short post mortem in the releasing, ci & RTA team as well and one big thing there was those late changes, both in qt and installer side
[17:12:41] <thiago> it will also catch "we missed QtDBus on Mac" if it happens again
[17:13:32] <kkoehne> jaheikki3: Question is whether there where 'late changes' that weren't necessary, or whether they where issues that should have been fixed earlier?
[17:14:10] <fkleint> yep, more reliable RTA testing and finally when build failed -> exit 1 and discard packages to avoid testing all these broken packages..
[17:15:05] <jaheikki3> kkoehne: I think there were both :( As usual RC and "final" packages got most testing effort & some issues were found too late
[17:15:36] <jaheikki3> kkoehne: and some changes were done & take in even those weren't that critical
[17:17:02] <jaheikki3> fkleint: that is pretty much true. Only problem is that also RTA is taking some time and as you know we have always a bit hurry... That's why test request are sent before we have RTA results available
[17:17:42] <jaheikki3> fkleint: But we are trying to make it better all the time...
[17:18:37] <thiago> any suggestions to avoid this?
[17:19:02] <fkleint> jaheikki3: What is the status with new CI? As I understood, it was supposed to create packages
[17:19:13] <fkleint> jaheikki3: so, they should fall out of it automagically?
[17:19:22] <jaheikki3> thiago: what do you mean?
[17:20:46] <jaheikki3> fkleint: Well, the goal is to get binaries from new CI as well but let's see, we are starting with old packaging system & taking binaries from new CI when all ready
[17:21:56] <thiago> is it going to be ready by the beta?
[17:21:59] <thiago> we don't need it for the alpha
[17:22:22] <jaheikki3> thiago: Hoping so but I am not sure
[17:23:03] <thiago> I really think we need this
[17:23:13] <thiago> we aren't running the old CI with 5.6, so we don't know the state
[17:23:22] <thiago> if we have to depend on it for beta binaries, we risk serious delays
[17:24:02] <thiago> anyway, on the "suggestions to avoid this", I was wondering if we had a suggestion to avoid late changes that broke stuff.
[17:24:09] <thiago> we need to be stricter on the branch
[17:24:13] <thiago> we were too lax in 5.5
[17:24:17] <thiago> myself included
[17:24:31] <jaheikki3> thiago: We don't. We will have old packaging system up and running as long as new system works reliable
[17:24:43] <thiago> for 5.6, I am worried about the turning on of C++11 by default. This may have far-reaching consequences.
[17:25:39] <carewolf> I doubt it. It only affects Linux, right? MSVS has default c++11 and so does Apple clang?
[17:26:16] <thiago> OS X too
[17:26:21] <thiago> this is for *user* code
[17:26:28] <thiago> so it affects everything except MSVC
[17:26:33] <carewolf> right
[17:27:11] <carewolf> I thought clang was default c++11. G++ 5.2 will be default c++14, so projects needs to adapt anyway
[17:27:39] <carewolf> no not 5.2.. New number makes that g++ 6
[17:27:50] <thiago> that's 6.0, not 5.2.
[17:28:31] <thiago> glad that they decided to announce that chnage before ours, since it gave me the time to rework so I don't override the C++14
[17:28:43] <jaheikki3> thiago: Yeah, I agree we need to be stricter on release branch, we all. I am hoping all developers starts being stricter as well as we who stage changes
[17:29:03] <thiago> jaheikki3: also on the 5.6 branch
[17:29:21] <jaheikki3> thiago: True
[17:30:01] <thiago> are we creating binaries now, already, for beta?
[17:30:09] <thiago> if not, how long before the release are we planning to?
[17:30:19] <carewolf> you mean alpha?
[17:30:22] <kkoehne> thiago: Btw, is there a change also for cmake to make c++11 the default?
[17:30:43] <thiago> from my outside perspective, the 5.5 releases appeared to be delayed because we hadn't been creating the binaries so we didn't know how well they were building
[17:31:16] <thiago> kkoehne: cmake should do it on its own. This was a qmake-wide change (default_post.prf), so not specifically to Qt.
[17:31:34] <thiago> carewolf: no, alpha has no binaries. I meant if we've begun to work on the beta binaries already.
[17:31:48] <kkoehne> thiago: okay
[17:31:49] <jaheikki3> thiago: Yes, we have started to create binaries but at the moment qt5.git content is quite old... Hoping we will get it updated soon & after that first binary snapshot should be available quite soon
[17:32:14] <thiago> ok, here's a question then: should we ditch the qt5.git linking?
[17:32:41] <thiago> or is it a necessary part of QA? What do you think?
[17:32:58] <kkoehne> I think we agreed to already dropped the testing part in qt5.git
[17:33:05] <kkoehne> I mean the autotests?
[17:34:27] <jaheikki3> New ci will handle that integration differently, but implementation is still ongoing. If I have understood plans correctly there won't be autotest anymore at final qt5.git integration, just compile check
[17:35:05] <thiago> I meant waiting for qt5.git to integrate to create packages for a release
[17:37:43] <kkoehne> thiago: +1 Would require though that we explicitly log the submodule SHA's somewhere in the package
[17:38:02] <jaheikki3> Well, I don't have so strong opinion for it. For me it is quite clear way to produce packages and if new ci will make it much easier/reliable I think it should be OK teh
[17:39:30] <jaheikki3> lets see what new CI does & if there is still issues with it then discuss how to proceed...
[17:40:56] <jaheikki3> Is there something else for 5.5 post mortem or should we move to 5.6 packaging changes?
[17:41:58] <fkleint> 5.5.1 plans would be interesting?
[17:42:57] <jaheikki3> fkleint: We shortly discussed that in last meeting: branching ~end of august, 5.5.1 release during September
[17:43:28] <fkleint> oki good [ assuming CI does not completely fall apart ]
[17:44:08] <jaheikki3> fkleint: 5.5.1 will use old CI, should work (tm) ;)
[17:44:43] <jaheikki3> Ok, then 5.6 packaging related changes:
[17:46:45] <jaheikki3> I think with MacOS we might be able to use same packaging system, sw & binary installers as with 5.5 or does someone disagree?
[17:46:52] <thiago> )!(@#)(! hotel wifi...
[17:48:47] <jaheikki3> With windows we need to start using MSVC 2015 so I propose to replace MSVC 2010 & MSVC 2012 binary installers with MSVC 2015 ones, meaning we will have MSVC 2013 and MSVC 2015 installer for 32 and 64 bit windows
[17:49:02] <thiago> we will drop 2010, period
[17:49:05] <thiago> it won't compile
[17:50:07] <jaheikki3> I don't want to increase amount of installers so I hope we can drop 2012 as well
[17:50:38] <jaheikki3> If someone wants to use it he can compile qt by him/her self
[17:51:13] <thiago> ok
[17:51:18] <jaheikki3> And with Linux I am wondering if we could finally drop 32 bit isntallers as well?
[17:52:06] <jaheikki3> any opinions about that?
[17:53:08] <thiago> what are the download rates?
[17:53:51] <jaheikki3> akseli: do you have some statistic from 5.5 yet?
[17:56:14] <akseli> jaheikki3: not that specific statistics yet
[17:56:16] <jaheikki3> with 5.4 it was a bit more than 10% if I remember correctly
[17:57:25] <thiago> I wonder how many of those are erroneous downloads
[17:57:33] <thiago> I propose we drop it for the beta and we'll see if anyone complains
[17:57:56] <jaheikki3> thiago: +1
[18:00:17] <jaheikki3> Ok, it seems no one disagree so lets try that.
[18:00:57] <carewolf> +1
[18:01:07] <akseli> +1
[18:01:45] <jaheikki3> I think this was all at this time. I think we can skip next week & have new meeting 1.9 at this same time. Then we should have knowledge if Alpha can be released as planned
[18:01:50] <jaheikki3> OK?
[18:02:31] <thiago> wait, what?
[18:02:37] <thiago> shouldn't the alpha come out this week or the next?
[18:02:40] <thiago> what's holding it up?
[18:03:10] <jaheikki3> https://wiki.qt.io/Qt-5.6-release :Alpha 8 Sep
[18:03:23] <thiago> the date sounds wrong
[18:03:27] <thiago> why is it so late?
[18:03:31] <thiago> we're branched
[18:04:05] <thiago> or, asked differently: what do we need to do before we release the alpha?
[18:04:13] <thiago> why can't we build tarballs now?
[18:04:36] <jaheikki3> Get latest content compile together & in qt5.git
[18:04:49] <jaheikki3> At the moment there is issues in CI
[18:05:09] <thiago> let's try to release ASAP, even if it's before that date
[18:05:25] <thiago> I see no reason we need 3 weeks for it
[18:05:58] <thiago> we release the first qt5.git integration that builds
[18:06:03] <thiago> agreed?
[18:06:05] <jaheikki3> Yeah, of course. I think we will release alpha when we can get latest stuff in qt5.git & it can be compiled
[18:06:20] <thiago> the first after today
[18:06:36] <jaheikki3> Ok for me
[18:06:55] <thiago> ok, good
[18:07:13] <fkleint> crap, looks like I missed a quite part due to network hickup
[18:07:58] <thiago> fkleint: yeah, I did too, ebfore
[18:08:14] <fkleint> can someone please paste it? or send conclusions
[18:08:18] <thiago> we agreed to drop the Linux 32-bit installer for the beta and see if anyone complains
[18:08:25] <fkleint> was there sth on this VS2010/WEC7 issue?
[18:08:33] <thiago> we agreed to drop MSVC 2010 (not supported anyway) and MSVC 2012 (to make room for 2015)
[18:08:42] <fkleint> ok, but what about VS2008
[18:08:45] <fkleint> for WEC7
[18:08:46] <fkleint> ?
[18:08:46] <thiago> not supported
[18:08:52] <fkleint> in 5.6?
[18:08:53] <thiago> they will *stop* *compiling*
[18:08:59] <thiago> wait, wait
[18:09:03] <thiago> I'm getting ahead of myself
[18:09:06] <thiago> that's 5.7!
[18:09:13] <thiago> scratch all of that
[18:09:18] <fkleint> exactly - 5.6 is the LTS release that is suposed to  have WEC 7
[18:09:27] <thiago> we agreed to drop 2010 and 2012, to make room for 2015
[18:09:34] <thiago> they will still build if you want, manually
[18:09:42] <fkleint> (personally would love to get rid of it , but Gunnar & kdab will go ballistic)
[18:10:08] <fkleint> as for Desktop packages, yes makes sense 13 + 15
[18:10:25] <thiago> and we agreed to release 5.6 alpha the first opportunity we get: the first qt5.git integration that passes after the necessary merges have happened
[18:10:48] <thiago> that should hopefully take less than 3 weeks
[18:11:21] <fkleint> Ah,ok
[18:12:19] <jaheikki3> And next meeting will have after 2 weeks, at this same time
[18:12:28] <fkleint> good
[18:13:18] <jaheikki3> I need to go now so let's end this meeting. I'll sent memo & invitations tomorrow, thanks & bye!
[18:13:31] <fkleint> bye
[18:13:57] <thiago> thanks jani


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.qt-project.org/pipermail/releasing/attachments/20150819/7b96f20c/attachment.html>


More information about the Releasing mailing list