[06:37] <alf_> RAOF: Hi! I was playing with debian/control to solve https://bugs.launchpad.net/mir/+bug/1348515 and noticed some interesting results.
[06:37] <alf_> RAOF: If we remove the version from Breaks,Replaces then the replacement doesn't happen and the builds fail.
[06:37]  * RAOF wonders again why we have a mircommon-dev :)
[06:38] <alf_> RAOF: If we set the version to something less strict (e.g. << 0.6) then both packages get installed :/
[06:38] <alf_> RAOF: Conflicts,Replaces,Provides works as expected
[06:38] <RAOF> Is something in the CI infrastructure manually install mircommon-dev?L
[06:39] <RAOF> Anything with Provides: is incorrect.
[06:39] <RAOF> mircommon-dev isn't a virtual package, with multiple different implementations that you can select from :)
[06:40] <alf_> RAOF: @mircommon-dev installed in CI, Yes, we fixed that, but that's ortogonal to "New mircommon-dev is not replaced by libmircommon-dev"
[06:40] <RAOF> Ah, so we just need to fix the versioning.
[06:40] <RAOF> Which our processes make unnecessarily difficult, because we don't *have* a version.
[06:40] <RAOF> (This is why the packaging branch should not be trunk)
[06:41] <RAOF> :)
[06:41] <alf_> RAOF: right, but changing the version to e.g. << 0.6, doesn't work well both packages get installed :/
[06:42] <alf_> RAOF: According to https://wiki.debian.org/Renaming_a_Package , the way to do it is either Conflicts,Replaces,Provides (and it doesn't have to be a virtual package), or Breaks,Replaces + transitional package
[06:42] <RAOF> Because the package version is now 0.6.0bzr1792pkg0utopic5+autopilot0, and 0.6.0bzr is not strictly less than 0.6
[06:43] <RAOF> We don't need a transitional package because nothing that isn't in src:mir depends on mircommon-dev
[06:44]  * RAOF will be back soon
[06:46] <alf_> RAOF: @"0.6.0bzr is not strictly less than 0.6", I don't think that's the reason, our latest mircommon-dev is still 0.5.0.
[06:47] <alf_> RAOF: https://jenkins.qa.ubuntu.com/job/mir-mediumtests-runner-mako/2209/console , is a run with Breaks and Replaces with version (<< 0.6~) .
[06:48] <alf_> RAOF: It seems we are setting up both packages: "Setting up libmircommon-dev:armhf (0.6.0bzr1797pkg0utopic24+autopilot0)" "Setting up mircommon-dev:armhf (0.5.0+14.10.20140724-0ubuntu1)" which is weird
[06:59] <RAOF> Hm.
[06:59] <RAOF> I shall check exactly how apt invokes dpkg.
[07:02] <RAOF> Hm. Actually, why not solve the problem once and for all...
[07:02] <RAOF> alf_: Do you know how Alan fiddled with the setup process?
[07:02] <RAOF> alf_: Rather than calling dpkg exactly the way apt would call it, it occurs to me that it'd be simpler to just call apt :)
[07:04] <alf_> RAOF: AFAIK, the runner scripts are here: https://code.launchpad.net/~mir-team/+junk/mir-medium-test-runner-for-jenkins/
[07:06] <RAOF> Sweet.
[07:06] <RAOF> Let's play “local apt repository” then...
[07:08] <alf_> RAOF: Note that if you want to try debian/control fixes, base them against lp:~afrantzis/mir/fix-mircommon-debian-replaces, since in this branch I have reverted the fix for not installing mircommon-dev which is now in lp:mir/devel. If you just use mir/devel you will not get mircommon-dev at all in CI so any fixes will not have an effect.
[07:09] <RAOF> Really? I thought it pulled mircommon-dev from the apt-get install glmark2-es2-mir?
[07:09] <RAOF> (ie: from the package in the archive)
[07:16] <RAOF> Ok. I think that *should* work. I might wait until Alan is on and ask him how to test it :)
[07:16] <RAOF> But first, some shopping.
[14:33] <racarr_> GOod morning...probably missing standup...sorry..have to go out and get a USB Key accidentally deleted my refit partition (meant to just delete OSX)
[14:33] <racarr_> and now computer doesnt boot :)
[18:19] <racarr_> I have never understood
[18:19] <racarr_> wqhat the Native
[18:19] <racarr_> in NativePlatform
[18:19] <racarr_> means
[18:39] <kdub> racarr_, it is confusing
[18:39] <kdub> we really have buffer-allocating platform code and display-showing platform code
[18:40] <kdub> and we awkwardly mix it into that class for the modes that don't need the display-showing code
[18:45] <racarr_> kdub: I guess I understand the purpose I just odnt understand
[18:46] <racarr_> the word "native"
[18:46] <racarr_> I guess its kind of like
[18:46] <racarr_> for a nested server
[18:46] <racarr_> its certain hooks to the "underlying native" platform or something
[18:46] <racarr_> I dunno
[18:47]  * kdub thinks the structure is a bit wrong, and the name is a bit of a bandaid
[18:47] <kdub> but, bigger fish to fry
[18:48] <racarr_> all I know is LUNCH LUNCH LUNNNNNNNNNNNNNNNNCH
[18:48] <racarr_> kdub: Indeed. I was only bringing it up incase there was some like
[18:48] <racarr_> clever way of looking at it I was missing
[18:48] <racarr_> lol
[19:27] <racarr_> Back