[00:15]  * josharenson is all ears
[00:19] <josharenson> ack
=== nudtrobert1 is now known as nudtrobert
[09:01] <Saviq> moin
[09:03] <Saviq> tsdgeos, linker Q... given libraries A and B, B referencing a symbol in A, would you expect an "undefined reference to..." error to depend on what flags you build *B* with?
[09:03] <tsdgeos> Saviq: yep
[09:04] <tsdgeos> not all linking makes the linker complain about missing symbols
[09:04] <tsdgeos> sometimes it goes for "ok maybe we get that later"
[09:04] <tsdgeos> i think
[09:05] <Saviq> tsdgeos, case in point: Josh's greeter links to liblightdm-qt5-3, builds fine everywhere, except on xenial when CMAKE_BUILD_TYPE=Rel*
[09:05] <Saviq> when it fails to find QLightDM::SessionsModel::roleNames(), which, btw, is not implemented in there, but SessionsModel inherits QAbstractItemModel
[09:06] <Saviq> so that's what it relies on elsewhere
[09:06] <tsdgeos> weird
[09:06] <Saviq> separate point is that it uses deprecated setRoleNames
[09:06] <tsdgeos> are we sure that's not a dirty builddir or something?
[09:06] <Saviq> yes
[09:07] <Saviq> tsdgeos, https://code.launchpad.net/~josharenson/unity8/sessions-model/+merge/278081/comments/707133 is where the failure first happened, then I managed to track it down to build type
[09:08] <Saviq> it fails in all builds on xenial there with "undefined reference to `QLightDM::SessionsModel::roleNames() const'"
[09:08] <Saviq> but builds fine on vivid, and when CMAKE_BUILD_TYPE=Debug
[09:08] <tsdgeos> maybe lightdm-qt is actually broken in xenial?
[09:08] <tsdgeos> is that the lightdm we build or the real thing?
[09:09] <Saviq> the real thing, built from lp:lightdm
[09:09] <Saviq> tsdgeos, fwiw readelf shows the QAbstractItemModel::roleNames symbol
[09:10] <Saviq> which ~makes sense since that's where it needs to come from, as QLightDM::SessionsModel does not implement it at all
[09:11] <Saviq> so yeah, well, a separate question altogether is whether it works or not (and that it should be updated to non-deprecated roleNames())
[09:12] <Saviq> but that seems to be a behaviour change at least between gcc 4.9 and 5.0 or something
[09:12] <Saviq> and I'm not totally sure which one's wrong :)
[09:12] <tsdgeos> that's weird yeah
[09:13] <Saviq> FWIW readelf on vivid shows the same symbols
[09:14] <tsdgeos> so this is only the session-chooser branch?
[09:14] <Saviq> sessions-model, yes, there's new code where Josh actually makes use of more of the QLightDM implementation
[09:15]  * Saviq makes a minimal example and files a bug against gcc
[09:15] <tsdgeos> Saviq: about setRoleNames it works fine from what i can see in the code
[09:17] <tsdgeos> hmm, but i had built the session-chooser branch
[09:17] <tsdgeos> i guess because i did a dewbug build?
[09:18]  * tsdgeos checks
[09:19] <Saviq> tsdgeos, yeah the default build is debug, try with ./build.sh Release
[09:45] <tsdgeos> Saviq: yeah can confirm, it's really weird, can't figure it out either what could be wrong
[09:46] <Saviq> tsdgeos, ok /me filing a toolchain bug
[09:46] <tsdgeos> yeah we'd need someone with more knowledge to have a look really
[10:01] <Saviq> greyback, what do you say we undo that one change in Daniel's branch for now, it wasn't necessary for fillMode to work was it?
[10:01] <Saviq> greyback, or at least modify it to put out the warning but still behave as before
[10:03] <Saviq> tsdgeos, the plot thickens, http://pastebin.ubuntu.com/13889040/ builds fine regardless
[10:04] <tsdgeos> Saviq: interesting
[10:04] <greyback> Saviq: modification would be possible, just lines 278-296 of the diff actually implement the feature
[10:05] <greyback> I've still not figured out root cause
[10:29] <Saviq> tsdgeos, d'oh!
[10:30] <tsdgeos> Saviq: what's wrong?
[10:32] <Saviq> tsdgeos, most likely something along the lines of http://paste.ubuntu.com/13889709/
[10:32] <tsdgeos> so the mocks are breaking stuff
[10:32] <tsdgeos> might be
[10:33] <Saviq> confirming now
[10:37] <Saviq> tsdgeos, yeah, that + a missing CMAKE_CURRENT_BINARY_DIR for the .moc, makes it build
[10:38] <Saviq> sessionbackend test fails, but that's an issue I'll leave to Josh
[10:38] <tsdgeos> Saviq: it's weird that only happens on release though :S
[10:40] <Saviq> tsdgeos, we do enable -Werror on release, but not in debug - while weird, it might make the difference (will confirm)
[10:47] <Saviq> greyback_, so wdyt, unblock silo 22 by going back to original behaviour of ignoring hasBuffer() for now?
[10:47] <greyback_> Saviq: it's what I'm testing now
[10:47] <Saviq> greyback_, ah koolz, tx
[10:47]  * Saviq looks sideways at all the branches will have to rebase
[11:16] <Saviq> tsdgeos, so yeah, only difference in the g++ command seems to be -O2 -DNDEBUG, I say the mocks get the symbol optimized away
[11:17] <tsdgeos> probably
[11:18] <Saviq> so it works in debug, even if linking against the wrong .so, but not in release
[11:19] <Saviq> and it "works", by chance, simply
[11:20] <Saviq> and in vivid it might not have been optimized away because with gcc 4.9 but is in xenial with 5.0
[11:21]  * Saviq happy it's not a toolchain issue, and that he's not completely dumb
[11:22]  * Saviq deserves food
[11:53] <mzanetti> wheee! I'm back
=== pat_ is now known as Guest15515
=== they_eye4 is now known as they_sad4
[15:31] <Saviq> @unity, dednick: standup
[15:31] <mterry> ltinkl, btw I rebuilt the oobe silo this morning
[15:31] <ltinkl> mterry, great thx
[15:41] <mzanetti> Saviq, there is an "Echo" option in mumble
[15:41] <mzanetti> it was set to "Mixed". I've set it to "Disabled" now. let's see how that goes :D
[15:41] <mzanetti> echo -> disabled
[15:41] <mzanetti> makes total sense
[15:48] <Saviq> mzanetti, and that's not echo cancellation?
[15:48] <mzanetti> it probably is
[15:49] <mzanetti> but setting it to anything else than "disabled" doesn't seem to help either
[15:49] <mzanetti> so...
[15:50] <Saviq> mzanetti, I think you were talking with Daniel about app resizing, the only artifact I can see now is that when resizing vertically, stuff jumps up/down, as if it was anchored to bottom left corner instead of top left?
[15:50] <mzanetti> hmm
[15:51] <Saviq> ok so maybe I'm wrong and you didn't talk about this, will have a chat when he's back
[15:51] <mzanetti> not this...
[15:52] <mzanetti> Saviq, but it's been ages since I tested/reviewed this stuff
[15:52] <mzanetti> iirc I did test that
[15:52] <mzanetti> might be a bad merge after the review
[16:03] <Saviq> mzanetti, so you mean you didn't see that behaviour?
[16:03] <Saviq> mzanetti, can you still check in silo 22?
[16:04] <Saviq> I won't block on that anyway as it's a big improvement, but we'll need to follow up
[16:04] <Saviq> so anyway, no need for you to test really
[16:04] <Saviq> mzanetti, ↑
[16:20] <tsdgeos> Mirv: any idea why the ubuntu-ui-toolkit package didn't migrate to main from proposed yet?
[16:21] <mzanetti> Saviq, ack (had to reboot)
[16:21] <Saviq> tsdgeos, DON'T!...
[16:21] <Mirv> tsdgeos: see any other channel like #ubuntu-devel. britney did interesting trade of keeping it in proposed due to s390x build issues. working on it.
[16:21] <Saviq> tsdgeos, britney is software, fwiw ;)
[16:22] <tsdgeos> ok
[16:22] <Saviq> tsdgeos, it's Qt 5.5, new poppler *and* new architecture all at the same time :)
[16:45] <mzanetti> Saviq, so you mean the content of the surface being jittery?
[16:45] <Saviq> mzanetti, yes, but more that its top-left corner isn't stab;le
[16:46] <mzanetti> Saviq, still content? or windeco itself?
[16:47] <Saviq> mzanetti, still content, say "System settings" in settings app
[16:55] <mzanetti> Saviq, ack, confirmed. will follow up on it
[17:29] <Mirv> Saviq: and new apt...
[17:29] <Mirv> which uncovered regressions in aptdaemon/python-apt/...
[17:29] <Mirv> all in a nice big bundle