| RAOF | <grayson-g[m]> "I've been looking at `boost::..." <- This may be a good time for you to look at adding the API alan_g referenced: [issue 2543](https://github.com/MirServer/mir/issues/2543)? | 00:00 |
|---|---|---|
| RAOF | Hm. Except you'd *actually* be wanting to add a timer, which would be exposing an `Alarm`-type API. | 00:01 |
| RAOF | But it's the same sort of thing. | 00:01 |
| -GitHub[m]:#mir-server- **[MirServer/mir]** bors[bot] merged [pull request #2601](https://github.com/MirServer/mir/pull/2601): Ensure test_wayland_generator runs (and fix test) | 06:12 | |
| -GitHub[m]:#mir-server- **[MirServer/mir]** bors[bot] closed [issue #2597](https://github.com/MirServer/mir/issues/2597): test_wayland_generator fails (but not run by ptest target) | 06:12 | |
| === dot-tobias is now known as dot-tobias[away] | ||
| RAOF | Hey hey Saviq. WRT the fact that [we don't test the *tarballs* we release](https://github.com/MirServer/mir/issues/2596), could we in fact have a github action that runs on final release, runs `tools/make_release_tarball`, and reports any failures? | 07:57 |
| RAOF | (We would also want to add a "run the tests" check to `make_release_tarball`, but that's easy) | 07:58 |
| Saviq | We really should have a pre- job that builds the tarball and source packages, where applicable, and goes from there | 07:58 |
| Saviq | Rather than every job checking out on its own | 07:59 |
| RAOF | Aaah, yes. That would be neatm. | 07:59 |
| RAOF | s/neatm/neat/ | 07:59 |
| Saviq | I'll put that as part of my release process rework | 08:00 |
| Saviq | We broked something: | 08:15 |
| Saviq | https://launchpad.net/~mir-team/+archive/ubuntu/dev/+packages?field.name_filter=mir&field.status_filter=&field.series_filter= | 08:15 |
| Saviq | It's the release merge that triggered it | 08:18 |
| === dot-tobias is now known as dot-tobias[away] | ||
| === dot-tobias[away] is now known as dot-tobias | ||
| alan_g[m] | Just seen the email storm. Are you looking into it? | 08:31 |
| Saviq | Nothing obvious so far | 08:32 |
| Saviq | The version bump? | 08:33 |
| alan_g[m] | unable to open lockfile /run/user/2001/wayland-0.lock check permissions... (full message at https://libera.ems.host/_matrix/media/r0/download/libera.chat/b782fab1ffacb67ca4786e13633333b3b72fb395) | 08:33 |
| alan_g[m] | That's how the tests are breaking. Now for why... | 08:34 |
| alan_g[m] | Saviq: Has the Launchpad infrastructure changed? | 08:37 |
| Saviq | Doubt that | 08:40 |
| Saviq | I will push the last working one with a version bump to confirm | 08:41 |
| alan_g[m] | Ack. My first suspicion is that a useless XDG_RUNTIME_DIR has been added to the environment | 08:42 |
| alan_g[m] | Because we have a hack along the lines of "if XDG_RUNTIME_DIR is missing, set it (to /tmp)" | 08:43 |
| alan_g[m] | Yup, we've now got the following user environment: | 08:51 |
| alan_g[m] | XDG_RUNTIME_DIR=/run/user/2001 | 08:51 |
| alan_g[m] | XDG_SESSION_CLASS=background | 08:51 |
| alan_g[m] | XDG_SESSION_ID=c80 | 08:51 |
| alan_g[m] | XDG_SESSION_TYPE=unspecified | 08:51 |
| alan_g[m] | Which is new with the failing builds | 08:51 |
| alan_g[m] | So the question is "what is wrong with `/run/user/2001` - does it exist? Etc?" | 08:53 |
| Saviq | I feel like we've been through this before... | 08:55 |
| alan_g[m] | Can you get access to an instance to poke around? | 08:55 |
| === dot-tobias is now known as dot-tobias[away] | ||
| === dot-tobias[away] is now known as dot-tobias | ||
| alan_g[m] | We have | 08:56 |
| Saviq | For reference, reported the issue to Launchpad: | 09:05 |
| Saviq | https://answers.launchpad.net/launchpad/+question/702995 | 09:05 |
| alan_g[m] | OK, I'll leave you with this one | 09:07 |
| -GitHub[m]:#mir-server- **[MirServer/mir]** Saviq requested a review from AlanGriffiths for [pull request #2603](https://github.com/MirServer/mir/pull/2603): tests: fix XDG_RUNTIME_DIR handling | 09:17 | |
| -GitHub[m]:#mir-server- **[MirServer/mir]** Saviq opened [pull request #2603](https://github.com/MirServer/mir/pull/2603): tests: fix XDG_RUNTIME_DIR handling... (full message at https://libera.ems.host/_matrix/media/r0/download/libera.chat/62bc3aaa1d4cd920f95e15dd759d7c8c1ee506b5) | 09:17 | |
| alan_g[m] | Saviq is there any pattern to your "transparent windows" issue? I've been working on my oldest laptop without seeing it yet. Is there a particular app it happens to? | 10:19 |
| Saviq | If I had to guess, it'd be Firefox affected most often, but that's mostly because that's what I'm looking at the most | 10:20 |
| Saviq | s/mostly/likely/ | 10:20 |
| alan_g[m] | OK, so I've got glmark running behind clion and firefox and seeing nothing. (yet) | 10:21 |
| Saviq | It does only happen occasionally, a few flashes in quick succession, then it's back in order again | 10:22 |
| Saviq | Between 1-3 times a day | 10:22 |
| alan_g[m] | OK, so may not show in an hour or two | 10:23 |
| -GitHub[m]:#mir-server- **[MirServer/mir]** bors[bot] merged [pull request #2603](https://github.com/MirServer/mir/pull/2603): tests: fix XDG_RUNTIME_DIR handling | 11:07 | |
| === dot-tobias is now known as dot-tobias[away] | ||
| === dot-tobias[away] is now known as dot-tobias | ||
| Saviq | alan_g ah, got an idea about my copy/paste issues. There's two c/p modes are there not? "online", where clients negotiate the data, and "cached", where it's a one way thing. So the latter is what's not working I think. If I quit the "source", I never get to paste anything | 12:18 |
| Saviq | Maybe Firefox / Thunderbird behave similarly to that with some copies (from the address bar, or through "copy link")? | 12:19 |
| alan_g[m] | There's only the one way we support, and that requires both source and destination to be alive | 13:11 |
| alan_g[m] | primary-select (which we don't support) could allow stuff to be cached by a third party | 13:14 |
| Saviq | Sure, but I don't mean primary-select. I mean Ctrl+C, close the source, Ctrl+V | 13:17 |
| Saviq | That, for at least some cases, should work | 13:17 |
| alan_g[m] | The source needs to hang around (possibly without a window) to negotiate mime type and data | 13:17 |
| alan_g[m] | The Mir code is pretty simple (just relaying between source and sink) if you want to poke around:src/server/frontend_wayland/wl_data_device.cpp | 13:22 |
| Saviq | Sure, I just mean we may need to complicate it to support this (valid, IMO) use case - regardless of whether this is actually what Firefox / Thunderbird is actually impacted by | 13:31 |
| alan_g[m] | You mean adding a new protocol and hoping clients implement it? | 13:31 |
| Saviq | I don't think there's now protocols needed. I mean, other compositors do it already | 13:33 |
| alan_g[m] | Other compositors implement primary selection | 13:34 |
| Saviq | Conceptually, Mir could take over the mime type negotation in the case when source went away, with the information it has | 13:34 |
| Saviq | But Ctrl+V doesn't paste primary selecton | 13:34 |
| Saviq | +i | 13:34 |
| Saviq | I'll have to log out into GNOME to confirm, but I believe the compositor takes over the mime type negotation if the source went away, with the information it does have | 13:35 |
| Saviq | It may not work always, but for simple cases like text the compositor can certainly do it | 13:35 |
| Saviq | Looks like the my fix didn't work... | 13:43 |
| alan_g[m] | Agreed | 13:43 |
| Saviq | In other news, https://bugs.launchpad.net/ubuntu/+source/gcc-12/+bug/1983852 spread to other architectures, and the breaking gcc-12 now migrated to the release pocket. | 13:44 |
| alan_g[m] | Feels like "Whac-A-Mole" | 13:47 |
| alan_g[m] | I should think before approving! You need something like `XDG_RUNTIME_DIR=/tmp GTEST_OUTPUT=xml:./ dh_auto_build -- ARGS="-V" ptest` | 14:08 |
| alan_g[m] | This is running in make, not bash | 14:08 |
| alan_g[m] | Shall I just push to main? | 14:09 |
| Saviq | Hmm but export should work in make? But maybe I indented and that was wrong? | 14:11 |
| Saviq | Right, it'd have to be `XDG_RUNTIME_DIR := /tmp\nexport XDG_RUNTIME_DIR` | 14:12 |
| alan_g[m] | No make basically runs the export command on a shell of its own | 14:12 |
| Saviq | That's with the indent, without the indent, that's a make command, not a shell one? | 14:12 |
| Saviq | I. Hate. Make. | 14:13 |
| alan_g[m] | Resistance is futile | 14:13 |
| alan_g[m] | Anyway, shall I push the fix? | 14:14 |
| alan_g[m] | nm done it. What can go wrong? | 14:20 |
| alan_g[m] | Saviq I've not seen your "transparent windows" issue despite looking for it all day. | 16:55 |
| alan_g[m] | Nor have I come up with a plausible theory to investigage (yet). | 16:56 |
| Saviq | Sorry :/ | 16:56 |
| alan_g[m] | I have, however noticed that with a nested session I can get garbled window content with windows spanning displays. (But you have see-through, not just garbage, so I think unrelated) | 16:58 |
| alan_g[m] | So, I think there's something wrong in the buffer management. And that might be a cause for both symptoms | 16:59 |
| alan_g[m] | Anyway, signing off in the hope of progress tomorrow | 17:01 |
| alan_g[m] | đŸ‘‹ | 17:01 |
| Saviq | My issues are likely to be gone tomorrow | 17:01 |
| Saviq | Just got the new laptop, so if it's load related… | 17:01 |
| alan_g[m] | You'll have the one where it manifests though. | 17:03 |
| Saviq | So I should not burn it on the stake? | 17:06 |
| Saviq | Good night o/ | 17:06 |
| -GitHub[m]:#mir-server- **[MirServer/mir]** wmww opened [issue #2604](https://github.com/MirServer/mir/issues/2604): Popups dismissed by Mir when parent clicked... (full message at https://libera.ems.host/_matrix/media/r0/download/libera.chat/dc6cabf51301d1dbea557642bceed1d7ec0edaea) | 18:28 | |
| -GitHub[m]:#mir-server- **[MirServer/mir]** AlanGriffiths opened [issue #2605](https://github.com/MirServer/mir/issues/2605): Nesting Wayland compositors should use pointer confinement... (full message at https://libera.ems.host/_matrix/media/r0/download/libera.chat/49ef59251f63d6ca3414f21a0b163bf264890be7) | 22:26 | |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!