dufluRAOF: Did you notice bug 1660017 ?06:40
ubot5bug 1660017 in Mir "EDID does not change when hotplugging a monitor" [Medium,New] https://launchpad.net/bugs/166001706:40
dufluI know it's almost EOD06:40
RAOFI did not.06:42
duflu(assuming hotplugging does not crash the server, which still happens but I have not clearly described in any recent bug)06:46
RAOFI am confused.06:53
RAOFWe're using the functions documented as doing a full hardware probe.06:54
alan_gduflu: as you've poked around Xmir, maybe you know the answer... I'm trying to use it in a snap, (where nothing is ever on the default path) and it can't find sxkbcomp. I can see it holds the format string "%s%sxkbcomp" -w %d %s -xkm "%s" -em1 %s -emp %s -eml %s "%s%s.xkm", but is there a way to set the first %s field? (A quick look at the code suggests it's a build-time constant, but I'm hoping I missed something.)09:43
duflualan_g: Did you build Xmir or copy from archive?09:45
alan_gfrom archive09:46
duflu(The trick to avoid similar issues when building Xmir is to configure with --prefix=/usr)09:46
alan_gduflu: it needs to be (at runtime) %SNAP/usr09:48
duflualan_g: No idea then sorry09:48
alan_gBut I guess you've confirmed that it's set at build time09:48
duflualan_g: Maybe this is the same problem in reverse. You need to build Xmir with --prefix=<replacement for /usr>09:48
alan_gAt build time we don't know %SNAP09:49
dufluYep. Xorg code is a bit older than that09:49
* duflu looks at the source09:51
duflualan_g: Looks easily configurable. The first %s is the bin dir and the second is forward or backslash09:53
dufluI could do that.09:53
alan_gack, But it isn't a priority (ATM), and likely not the last blocker.09:55
duflualan_g: Are you suggesting the first %s is empty or non-empty and wrong?09:56
dufluLooks like building with empty should work (and default to $PATH)09:57
alan_gNo, the first % is /usr and is wrong in a snap09:57
alan_git needs to be (at runtime) %SNAP/usr09:57
duflualan_g: Yeah looks like the source is written to nicely handle NULL and search the path, but our builds are hardcoded09:57
alan_gI was hoping that I'd missed an env variable or similar (windows has winFixupPaths()).09:59
duflualan_g: Sounds easy to test and fix. I'll just stop using the --prefix workaround and see what works then. Sadly there is no environment. It's popen'd with an optional hardcoded bin path09:59
alan_gthat's what I thought. :(10:00
dufluIt stumped me too. When I first looked at Xmir it didn't work because the default build uses /usr/local for it which is wrong on Ubuntu10:01
dufluI never changed any release build parms, just have a technique for building it in development.10:01
duflualan_g: Tomorrow. Now; dinner10:02
alan_gduflu: Have a good one!10:02
=== hikiko is now known as hikiko|ln
alan_ggreyback: starting to think about D&D support. Where are the "drag start, drag move & drop events" initiated (a toolkit API mir_window... call?) handled (a server interface?) and delivered (a toolkit MirWindowEventCallback event?)11:51
greybackalan_g: to begin, have you seen this doc? https://docs.google.com/document/d/1IztlvbrJHmvLe3im1xe5WIKHkh4-SlBguokBCgXMgLo/edit#11:56
alan_ggreyback: was skimming docs but lacking overview. That's the starting point?11:57
greybackalan_g: IMO yeah. It was roughly what we agreed on at the sprint in Montreal11:58
alan_ggreyback: thanks, will read before bugging you again11:59
greybackalan_g: np11:59
* alan_g wonders how to represent "list of drag surfaces + origins"12:06
=== hikiko|ln is now known as hikiko
dandraderalan_g, does a mir_buffer_stream_swap_buffers_sync ever time out? I'm debbuging a situation where I have an internal client (mirclient run in a thread) gets stuck in mir_buffer_stream_swap_buffers_sync() forever while because server is unable to drop buffers from it. that all happens when I'm trying to close that client.12:53
alan_gdandrader: no timeout AFAIK12:55
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader
alan_ggreyback: quick one I see "A list of drag surfaces + origins" in one of the diagrams. Is this buffer stream(s) + hotspot(s)? Why is there a list?14:23
greybackalan_g: list because multiple items may be selected, and we thought it made sense to allow shell to decide how those items can be laid out14:24
greybackwe've not had much design input until recently14:25
greybackalan_g: this is all I know: https://docs.google.com/presentation/d/1NEcPwodymSAjvsUGWlDx7ocDLhs41edkPGq_MsA_BdM/edit#slide=id.g19dd5d1a07_2_2314:25
greybackalan_g: also note there is a process diagram in there, which differs to the one in the doc I gave you. The differences have yet to be figured out and clarified14:26
greybackI'm doing that right now actually14:26
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader
=== dandrader is now known as dandrader|afk
=== dandrader|afk is now known as dandrader

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!