| *** phoenix has quit IRC | 00:07 | |
| *** flatmush has quit IRC | 00:58 | |
| *** flatmush has joined #buildstream | 00:58 | |
| *** narispo has joined #buildstream | 03:00 | |
| *** narispo has quit IRC | 03:07 | |
| *** narispo has joined #buildstream | 03:07 | |
| *** juergbi has quit IRC | 06:19 | |
| *** juergbi has joined #buildstream | 06:25 | |
| *** rdale has joined #buildstream | 07:34 | |
| *** juergbi has quit IRC | 07:37 | |
| *** juergbi has joined #buildstream | 07:41 | |
| *** traveltissues has joined #buildstream | 08:46 | |
| *** bochecha has joined #buildstream | 09:16 | |
| *** phildawson has joined #buildstream | 09:26 | |
| *** phildawson_ has quit IRC | 09:31 | |
| *** phildawson_ has joined #buildstream | 09:32 | |
| *** santi has joined #buildstream | 09:45 | |
| coldtom | hi, bst seems to be unable to invoke `sh` inside the sandbox, even if i give an absolute path, anyone seen anything like this before? | 09:45 |
|---|---|---|
| Kinnison | is it in the sandbox? | 09:45 |
| Kinnison | is it dynamically linked to a libc which is absent? | 09:45 |
| benschubert | coldtom: Most of the time I had seen this is when I was missing a dependency on 'sh' | 09:46 |
| coldtom | it's definitely in the only dependency my element has | 09:46 |
| Kinnison | wrong arch? | 09:46 |
| benschubert | and if you run 'bst shell your_dependency' and run 'sh --help' does that work? | 09:46 |
| coldtom | it doesn't seem to find sh at all for that element, but with a `bst artifact checkout` /usr/bin/sh definitely exists | 09:47 |
| benschubert | can you chroot in it and sh --help ? | 09:48 |
| Kinnison | does /bin exist as a symlink to /usr/bin ? | 09:48 |
| Kinnison | Is the dynamic linker in place? | 09:49 |
| coldtom | ah /bin -> /usr/bin is not being created properly | 09:50 |
| *** jonathanmaw has joined #buildstream | 09:51 | |
| *** tiagogomes has joined #buildstream | 09:53 | |
| gitlab-br-bot | traveltissues opened issue #1188 (Check plugins conform to public API) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1188 | 09:57 |
| coldtom | hold up though, i modified the buildelement to make bwrap invoke /usr/bin/sh, surely that should have found it | 09:58 |
| benschubert | traveltissues: oh thanks for that | 09:58 |
| Kinnison | coldtom: were the other symlinks there? | 09:59 |
| traveltissues | sorry, benschubert i just saw your reply | 09:59 |
| traveltissues | feel free to amend the issue | 09:59 |
| benschubert | traveltissues: no worries :) thanks for adding the issue! | 09:59 |
| benschubert | nah, I believe it captures everything! | 09:59 |
| benschubert | I guess that means my MR is good to go? | 09:59 |
| coldtom | Kinnison: none of the / -> /usr symlinks got created properly, not sure on any others | 10:00 |
| gitlab-br-bot | traveltissues approved MR !1680 (bschubert/fix-git-ref->master: _gitsourcebase.py: Return 'None' when the ref is unset) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1680 | 10:00 |
| Kinnison | coldtom: then /lib -> /usr/lib was missing so ld-linux.so wouldn't be findable | 10:00 |
| persia | coldtom: Your /usr/sh might be linked to something in /lib, which it might not find in /usr/lib, etc. | 10:00 |
| Kinnison | Even your compiled programs need an interpreter | 10:00 |
| coldtom | ahhh ty | 10:03 |
| coldtom | now just to fix the missing symlinks | 10:04 |
| Kinnison | underlay underlay, amoeba amoeba, underlay! | 10:04 |
| * benschubert is lost | 10:05 | |
| * Kinnison apologises, 'tis a poor pun | 10:05 | |
| Kinnison | There was a cartoon mouse (Speedy Gonzales) who would shout "¡Ándale, ándale! ¡Arriba, arriba! ¡Ándale!" | 10:06 |
| * Kinnison imagines these days that'd be a fairly problematic thing for a children's cartoon | 10:09 | |
| benschubert | Remember this one, might be pronouncing incorrectly since I didn't make the link at all :'D | 10:14 |
| benschubert | traveltissues: I wonder whether BST_NO_PRESTAGE_KEY should not be something like 'BST_NO_REF' or something like that? | 10:14 |
| benschubert | since the result of this is that the source never gets a ref | 10:14 |
| traveltissues | benschubert, i'm not married to the name tbf, both you and juergbi seem to prefer that | 10:18 |
| traveltissues | it's true that it doesn't have a ref but it's also true that it doesn't have a key without staging | 10:19 |
| benschubert | traveltissues: ok :) I was just working on trying to remove the need to get the consistency in all cases not involving opening a workspace, and it seemed a bit more logical :) | 10:19 |
| benschubert | right | 10:19 |
| benschubert | I also wonder whether we'll ever have something like that that is not local files... | 10:19 |
| persia | benschubert: For the remote execution case, it might be an opportunity for acceleration to avoid having a local copy when it is known to exist remotely. | 10:21 |
| traveltissues | benschubert, do you mean local plugin or on disk files | 10:21 |
| juergbi | theoretically, something like a workspace shared across a network could be created | 10:22 |
| juergbi | I'd probably recommend a git repo instead but there might be a use case | 10:22 |
| coldtom | benschubert: is there a version of bst known to work with your bschubert/freedesktop-sdk-19.08.3-bst2 branch? | 10:23 |
| traveltissues | actually, thinking about it for a second, BST_NO_REF would probably be better | 10:23 |
| *** lachlan has joined #buildstream | 10:30 | |
| *** bochecha has quit IRC | 10:40 | |
| gitlab-br-bot | traveltissues closed issue #1184 (Remote Execution tests produce too much logs) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1184 | 10:42 |
| gitlab-br-bot | marge-bot123 closed issue #1184 (Remote Execution tests produce too much logs) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1184 | 10:43 |
| gitlab-br-bot | marge-bot123 closed issue #1184 (Remote Execution tests produce too much logs) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1184 | 10:43 |
| gitlab-br-bot | marge-bot123 merged MR !1677 (traveltissues/1184->master: buildgrid-compose.yml: reduce log level to WARN) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1677 | 10:43 |
| benschubert | traveltissues: on disk | 10:44 |
| benschubert | coldtom: master should if you remove the 'artifact' section in project.conf | 10:45 |
| tpollard | would be really nice to find that regression | 10:45 |
| benschubert | I personally don't have the bandwidth for this currently, but I'm happy to give pointers. tlater[m] might also have a few clues? | 10:46 |
| * tlater[m] doesn't know by heart what features it uses | 10:47 | |
| tlater[m] | coldtom: Have you tried any? I'd imagine master *should* be functional | 10:47 |
| benschubert | tlater[m]: the problem is about the artifact server not having its capabilities validated anymore before trying to contact it | 10:47 |
| benschubert | tlater[m]: master doesn't work because of the bug with the check of capabilities :) | 10:47 |
| coldtom | i'm using 1.91.2 atm, but importing the / -> /usr symlinks seems to be going wrong | 10:48 |
| traveltissues | :/ | 10:48 |
| tlater[m] | Ah interesting | 10:48 |
| tpollard | https://gitlab.com/BuildStream/buildstream/issues/1178 | 10:48 |
| benschubert | ouch that's bad | 10:48 |
| benschubert | mind opening an issue with all information on how to reproduce? :) | 10:48 |
| *** lachlan has quit IRC | 10:49 | |
| tlater[m] | coldtom: Could you point me to the exact element? | 10:51 |
| coldtom | tlater[m]: bootstrap/build/base-sdk/symlinks.bst as-is tells you that there's nothing staged in / | 10:52 |
| coldtom | i tried making empty directories for the symlinks we import to point at, but this just made the element have empty directories at /bin etc | 10:53 |
| tlater[m] | That's supposed to work | 10:53 |
| tlater[m] | Even more interestingly, I'm pretty sure I built this just the other day | 10:53 |
| tlater[m] | coldtom: Could you try actual master? | 10:53 |
| tlater[m] | We test that in CI, so it should be functional | 10:54 |
| traveltissues | what sort of sources are these btw, i have a sinking feeling | 10:54 |
| coldtom | local traveltissues | 10:54 |
| traveltissues | oh god | 10:54 |
| tlater[m] | traveltissues: local | 10:55 |
| tlater[m] | Did someone change the way we handle symlinks for local elements? | 10:55 |
| benschubert | tlater[m]: the freedesktop sdk tests in CI are broken for more than 2 weeks because of the capabilities bug (#1178) | 10:55 |
| tlater[m] | Oh, again :| | 10:56 |
| * tlater[m] thought we'd just fixed them | 10:56 | |
| coldtom | weird, i'm not getting that behaviour with 1.91.2 now :S | 10:57 |
| tlater[m] | traveltissues: Was that a mistyped "oh good"? | 10:57 |
| traveltissues | no | 10:57 |
| tlater[m] | You'll need to explain then! | 10:57 |
| benschubert | traveltissues: you mean you know what the problem is? | 10:57 |
| traveltissues | well, local sources are now advertising BST_NO_PRESTAGE_KEY, so they will be imported directly from cache in staging | 10:58 |
| benschubert | traveltissues, juergbi do we ever want to support a source not supporting CAS based directoy, which has no PRESTAGE key? (because that fails... interestingly currently) | 10:58 |
| *** lachlan has joined #buildstream | 11:02 | |
| *** phildawson_ has quit IRC | 11:02 | |
| juergbi | benschubert: no, the plan has been for a while to completely drop support for plugins that require non-virtual directory | 11:02 |
| juergbi | we should make sure we do this well before 2.0 | 11:02 |
| traveltissues | benschubert, i suspect that this might actually be an issue with symlinks on the buildbox side | 11:04 |
| traveltissues | but i'm not sure | 11:04 |
| traveltissues | this might be uncovered by the change to the new flag | 11:05 |
| tlater[m] | coldtom: The answer in that case is, use whichever version came before the BST_NO_PRESTAGE :) | 11:05 |
| * tlater[m] blames it to see which commit that is | 11:06 | |
| *** bochecha has joined #buildstream | 11:06 | |
| juergbi | traveltissues: the stage method of the local plugin also went through import_files and casd before the new flag, though | 11:07 |
| tlater[m] | coldtom: b447e9f5 is the last commit that should work | 11:07 |
| juergbi | not sure why the actual import would lead to different results | 11:07 |
| tlater[m] | Or b447e9f59537def81b24ab2e0e91e827d6f194bb if you want the full sha | 11:07 |
| traveltissues | juergbi, that is true. the only change really is that staging is being done earlier | 11:10 |
| traveltissues | and ofc the unique key | 11:11 |
| *** phildawson_ has joined #buildstream | 11:11 | |
| coldtom | ta tlater[m], i'll raise an issue in a bit | 11:11 |
| *** lachlan has quit IRC | 11:11 | |
| tlater[m] | coldtom: Would be nice if you could verify that commit works, too | 11:11 |
| coldtom | tlater[m]: i will take a look after this meeting :P | 11:12 |
| tlater[m] | ta :) | 11:12 |
| gitlab-br-bot | traveltissues opened MR !1681 (traveltissues/changeflag->master: Replace flag name) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1681 | 11:23 |
| benschubert | juergbi: ok great, I'll adapt our test plugins then! | 11:29 |
| gitlab-br-bot | marge-bot123 merged MR !1680 (bschubert/fix-git-ref->master: _gitsourcebase.py: Return 'None' when the ref is unset) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1680 | 11:32 |
| benschubert | Can anyone explain to me what is the purpose of the 'FetchSource' in tests/frontend/project/sources/fetch_source.py ? Because it really doesn't behave like any other kind of sources... | 11:43 |
| *** phildawson_ has quit IRC | 11:43 | |
| *** phildawson_ has joined #buildstream | 11:43 | |
| benschubert | (no refs, put a fetch phase) | 11:49 |
| *** lachlan has joined #buildstream | 11:54 | |
| traveltissues | coldtom, 66045d9c | 11:55 |
| traveltissues | this still fails to build that element, casd returns error 13 while trying to stage, which aiui is an access error | 11:56 |
| traveltissues | buildbox-common 04dd7473, buildbox-casd a389a20e | 11:57 |
| coldtom | traveltissues: using the latest commits in buildbox has got it working with that bst commit | 11:59 |
| * coldtom tries again with the tip of master | 11:59 | |
| traveltissues | thanks | 12:00 |
| *** lachlan has quit IRC | 12:00 | |
| coldtom | yup, looks like it was casd being out of date | 12:01 |
| coldtom | ta traveltissues, sorry for all the panic | 12:01 |
| traveltissues | np | 12:01 |
| tpollard | I think we need a better way of communicating the need for casd upgrades | 12:03 |
| tpollard | often when tests stop working for me the first thing I try is updating it | 12:03 |
| traveltissues | yes | 12:03 |
| tpollard | but, that's not very intuitive | 12:03 |
| coldtom | build bst with bst, run the tests in bst and tell everyone to use the same version of casd as you build in the bst project | 12:04 |
| traveltissues | i'm not sure how simple it would be to communicate that through buildstream though | 12:04 |
| benschubert | coldtom: the problem is notifying updates :) We could however have a real version stirng in casd and check that in buildstream | 12:05 |
| *** lachlan has joined #buildstream | 12:06 | |
| traveltissues | we can, the issue is when you have head changing on both projects the version isn't guaranteed to work unless you reversion every commit | 12:06 |
| coldtom | use `git describe` to get commits since last tag and use that as a minor version point? | 12:07 |
| coldtom | still not ideal though | 12:07 |
| traveltissues | do we test buildstream head against buildbox head for overnights? | 12:09 |
| *** bochecha has quit IRC | 12:11 | |
| benschubert | I don't think so, same version as normal tests | 12:11 |
| traveltissues | it would probably make sense to do as you suggested then benschubert and communicate non-supported versions to the user | 12:13 |
| traveltissues | if we don't already | 12:13 |
| *** santi has quit IRC | 12:13 | |
| benschubert | we don't yet | 12:14 |
| gitlab-br-bot | traveltissues opened issue #1189 (Check versions of buildbox) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1189 | 12:18 |
| *** lachlan has quit IRC | 12:18 | |
| *** phil has joined #buildstream | 12:35 | |
| *** phildawson_ has quit IRC | 12:35 | |
| tlater[m] | Are the runners down? | 13:08 |
| * persia perceives the current bst/casd version game as an incentive to preserve stability for some corners of the API | 13:09 | |
| tlater[m] | Of course the moment I start working on something that needs to be tested via gitlab ci the runners go down |: | 13:12 |
| *** bochecha has joined #buildstream | 13:16 | |
| tpollard | tlater[m]: dashboard for digitalocean has nothing screaming as of yet | 13:18 |
| *** santi has joined #buildstream | 13:19 | |
| tlater[m] | tpollard: https://gitlab.com/BuildStream/buildstream-docker-images/-/jobs/339464992 | 13:19 |
| tlater[m] | It's happening pretty consistently | 13:19 |
| * tpollard checks on the bastion | 13:19 | |
| * tlater[m] waits for the intermittent network failure to correct itself | 13:20 | |
| tpollard | I can connect to it's console, hopefully just some blips... | 13:21 |
| *** phoenix has joined #buildstream | 13:25 | |
| coldtom | anyone had any thoughts on making the git plugin work nicely with git-lfs? | 13:30 |
| tlater[m] | coldtom: I've never seen any discussion of that here | 13:32 |
| coldtom | there's https://gitlab.com/BuildStream/buildstream/issues/567 | 13:32 |
| coldtom | i hit the issue recently while trying to build fdsdk, icu uses git-lfs | 13:33 |
| tlater[m] | interesting | 13:33 |
| * tlater[m] suspects the discussion practically stopped there | 13:33 | |
| * coldtom too | 13:34 | |
| tpollard | tlater[m]: if it doesn't start working ping me and I'll ssh into it | 13:36 |
| tpollard | *the bastion | 13:36 |
| tlater[m] | ta tpollard :) | 13:36 |
| *** jonathanmaw has quit IRC | 13:56 | |
| *** phoenix has quit IRC | 14:13 | |
| *** akvilebirgelyte has quit IRC | 14:21 | |
| *** akvilebirgelyte has joined #buildstream | 14:24 | |
| gitlab-br-bot | traveltissues opened (was WIP) MR !1679 (traveltissues/splitdepkeys->master: Split dependencies in the cache key dict into strong and weak) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1679 | 14:30 |
| *** lachlan has joined #buildstream | 14:31 | |
| *** dtf has joined #buildstream | 14:31 | |
| *** dtf has joined #buildstream | 14:31 | |
| *** lachlan has quit IRC | 14:38 | |
| *** dtf has quit IRC | 15:02 | |
| *** dtf has joined #buildstream | 15:02 | |
| coldtom | hi, i'm trying to create a bazel plugin, but it's being a bit of a pain about fetching external deps in the build. I have created a source transform plugin which invokes `bazel fetch`, but it symlinks all over and breaks itself. I think i can fix it if i call fetch from the correct place in the sandbox. would bst have anything against me | 15:11 |
| coldtom | implementing some way to do source fetching inside of a sandbox? | 15:11 |
| *** dtf has quit IRC | 15:11 | |
| *** dtf has joined #buildstream | 15:11 | |
| benschubert | coldtom: when discussing with some of the Bazel engineers, they were telling me you could fetch evrything without problems and it shouldn't have effects on the system (with enough flags). I'm not sure breaking the sandbox for source fetching in the build is a good way to go :/ Do you have an example reproduction somewhere? | 15:13 |
| coldtom | benschubert: maybe i should look a bit further into flags then :/ i definitely don't want to break the sandbox in the build, i was hoping to do the fetch in a source plugin that does it all in a sandbox | 15:16 |
| coldtom | the main issues i was having was bazel tracking it's own installed tools (@bazel_tools, @local_config_platform) at fetch time, and then having broken symlinks when it came to build time | 15:17 |
| coldtom | every time i fixed up one symlink, more broken ones appeared | 15:19 |
| *** dtf has quit IRC | 15:19 | |
| benschubert | coldtom: what command are you running? | 15:20 |
| coldtom | benschubert: i was running `bazel fetch` but with more flags | 15:21 |
| benschubert | what flags? :) | 15:21 |
| *** dtf has joined #buildstream | 15:22 | |
| coldtom | --output_base=/not/home/.cache was my attempt to get things out of ~/.cache plus i had --repository-cache to make sure we could check the cache | 15:23 |
| coldtom | i couldn't find a way to change install_base, otherwise things would have been simpler | 15:24 |
| *** dtf has joined #buildstream | 15:24 | |
| *** dtf has joined #buildstream | 15:24 | |
| benschubert | have yout ry with --experimental_repository_resolved_file=<a string> too? | 15:24 |
| benschubert | I *think* it changes part of the behavior | 15:24 |
| coldtom | i'll give it a go! thanks :) | 15:26 |
| *** lachlan has joined #buildstream | 15:33 | |
| *** lachlan has quit IRC | 15:39 | |
| *** santi has quit IRC | 15:45 | |
| *** lachlan has joined #buildstream | 15:52 | |
| *** santi has joined #buildstream | 16:02 | |
| benschubert | https://gitlab.com/BuildStream/buildstream/blob/master/src/buildstream/_pipeline.py#L436-437 is it me or the message should be 'some workspaces could not be found but haven't been closed' ? | 16:08 |
| *** traveltissues has quit IRC | 16:15 | |
| gitlab-br-bot | BenjaminSchubert approved MR !1671 (chandan/remove-build-track->master: frontend: Remove tracking options from `bst build`) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1671 | 16:19 |
| *** lachlan has quit IRC | 16:59 | |
| *** bochecha has quit IRC | 17:00 | |
| *** tpollard has quit IRC | 17:04 | |
| *** lachlan has joined #buildstream | 17:08 | |
| *** lachlan has quit IRC | 17:11 | |
| *** lachlan has joined #buildstream | 17:14 | |
| *** lachlan has quit IRC | 17:22 | |
| *** tiagogomes has quit IRC | 17:25 | |
| *** phoenix has joined #buildstream | 17:34 | |
| *** lachlan has joined #buildstream | 17:38 | |
| *** santi has quit IRC | 17:42 | |
| *** rdale has quit IRC | 17:55 | |
| *** lachlan has quit IRC | 18:09 | |
| *** lachlan has joined #buildstream | 18:10 | |
| *** phoenix has quit IRC | 18:21 | |
| *** phoenix has joined #buildstream | 18:22 | |
| *** narispo has quit IRC | 19:21 | |
| *** narispo has joined #buildstream | 19:22 | |
| *** traveltissues has joined #buildstream | 19:34 | |
| *** traveltissues has quit IRC | 20:11 | |
| *** lachlan has quit IRC | 20:26 | |
| *** cs-shadow has quit IRC | 21:10 | |
| *** phoenix has quit IRC | 21:37 | |
| *** narispo has quit IRC | 22:50 | |
| *** narispo has joined #buildstream | 22:52 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!