| *** testy has quit IRC | 03:52 | |
| *** toscalix has joined #buildstream | 08:09 | |
| gitlab-br-bot | coldtom approved MR !1698 (jjardon/powerpc64le->master: buildstream/_platform/platform.py: Add ppc64el, powerpc64 and powerpc64le) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1698 | 08:43 |
|---|---|---|
| *** traveltissues has joined #buildstream | 09:06 | |
| *** tpollard has joined #buildstream | 09:12 | |
| *** benschubert has joined #buildstream | 09:14 | |
| *** rdale has joined #buildstream | 09:18 | |
| gitlab-br-bot | coldtom approved MR !1697 (chandan/no-command-general-elements->master: Indicate that compose & stack plugins do not run commands) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1697 | 09:23 |
| gitlab-br-bot | traveltissues approved MR !1694 (coldtom/allow-dev-shm->master: _sandboxbwrap.py: Create /dev/shm in the sandbox) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1694 | 09:30 |
| gitlab-br-bot | tpollard approved MR !1697 (chandan/no-command-general-elements->master: Indicate that compose & stack plugins do not run commands) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1697 | 09:34 |
| *** santi has joined #buildstream | 09:38 | |
| gitlab-br-bot | BenjaminSchubert opened (was WIP) MR !1691 (bschubert/fix-children-termination->master: scheduler.py: Prevent the asyncio loop from leaking into subprocesses) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1691 | 09:47 |
| *** tiagogomes has joined #buildstream | 09:50 | |
| *** tiagogomes has quit IRC | 09:53 | |
| *** tiagogomes has joined #buildstream | 09:57 | |
| gitlab-br-bot | traveltissues approved MR !1696 (bschubert/expand-path-configs->master: _remote.py: Expand user in certificates paths) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1696 | 09:57 |
| benschubert | traveltissues: do you believe tests for the expansion should be duplicated for source caches and RE ? It's the same code path AFAIK | 09:58 |
| traveltissues | it looks like a sufficient test benschubert | 10:00 |
| benschubert | awesome, to marge then, thanks! | 10:00 |
| coldtom | hi, i'm experiencing a bug where every invocation of bst fails with "Timed out waiting for buildbox-casd to become ready", anyone got any ideas on how to debug? | 10:01 |
| benschubert | coldtom: latest master? Which version of buildbox-casd? | 10:02 |
| benschubert | (i.e. when did you update buildbox-casd) | 10:02 |
| coldtom | i just installed buildbox-casd, i think the issue was specifying a remote cache (as removing the conf just made things work) | 10:03 |
| coldtom | as in, latest master of buildbox-casd | 10:04 |
| benschubert | that's weird | 10:04 |
| traveltissues | benschubert, would you mind having another look at !1689 | 10:04 |
| gitlab-br-bot | MR !1689: skip tracking elements without trackable sources https://gitlab.com/BuildStream/buildstream/merge_requests/1689 | 10:04 |
| benschubert | anything in ~/.cache/buildstream/cas/logs/ ? | 10:04 |
| benschubert | traveltissues: sure | 10:04 |
| traveltissues | ty | 10:05 |
| benschubert | traveltissues: are the benchmark up to date/worth redoing? | 10:05 |
| traveltissues | they're up to date | 10:05 |
| benschubert | awesome | 10:06 |
| juergbi | benschubert: hm, _ForkableUnixSelectorEventLoop also accesses private members and even _loop_factory is not public | 10:09 |
| juergbi | the existing close() method is not good enough? | 10:10 |
| juergbi | I guess it fails the 'is not running' test | 10:11 |
| benschubert | juergbi: calling the existing close method will actually unregister the fd, which will cause the parent's ioloop to break. | 10:12 |
| benschubert | And it also break the 'is not running' yep | 10:12 |
| juergbi | benschubert: unregister the fd in what way? | 10:12 |
| benschubert | juergbi: also, in the bug tracker they are speaking about forbidding 'fork' when an ioloop is present completely, so we have a bigger problem here potentially :) | 10:13 |
| benschubert | https://github.com/python/cpython/blob/master/Lib/asyncio/selector_events.py#L94 if that is called in the child, the parent ioloop dies | 10:13 |
| benschubert | because of https://github.com/python/cpython/blob/master/Lib/asyncio/selector_events.py#L276 | 10:14 |
| benschubert | https://github.com/python/cpython/blob/master/Lib/selectors.py#L308 I *think* is the culprit on that | 10:15 |
| juergbi | ah, and you work around that by replacing _selector | 10:15 |
| benschubert | correct | 10:15 |
| juergbi | have to think about it for a bit | 10:15 |
| benschubert | Sure, please do, I don't want us to rush on this MR given what it touches :'D | 10:16 |
| juergbi | benschubert: I'm wondering whether it might be more robust to avoid accessing these internals and rather call set_event_loop_policy() on fork, creating a new event loop policy instance | 10:21 |
| juergbi | in addition to set_wakeup_fd(-1) and probably resetting signal handlers | 10:21 |
| benschubert | I'm unclear around `set_wakeup_fd(-1)`, does it break all the signals handling or does it just create a new file and all previous handlers become moot, do you know? | 10:23 |
| juergbi | it stops Python from writing the signal number to that fd | 10:23 |
| juergbi | this is independent of other signal handler settings | 10:23 |
| juergbi | except that it only works if the signal handler is not set to SIG_DFL | 10:24 |
| juergbi | (asyncio installs a dummy handler for this purpose) | 10:24 |
| juergbi | and if the child process were to create its own async loop, that should just work (it will set its own, new fd) | 10:25 |
| benschubert | We could probably do: | 10:25 |
| benschubert | ``` | 10:25 |
| benschubert | # First reset all signals (moved from ChildJob to this part for that purpose) | 10:25 |
| benschubert | set_wakeup_fd(-1) | 10:25 |
| benschubert | old_policy = asyncio.get_event_loop_policy() | 10:25 |
| benschubert | new_policy = type(old_policy)() | 10:25 |
| benschubert | asyncio.set_event_loop_policy(new_policy) | 10:25 |
| benschubert | del old_policy | 10:25 |
| benschubert | ``` | 10:25 |
| benschubert | Would that be fine for you? | 10:25 |
| *** jonathanmaw has joined #buildstream | 10:26 | |
| benschubert | might indeed be more robust | 10:26 |
| juergbi | yes, I think I'd prefer that to accessing internals | 10:27 |
| juergbi | benschubert: but we could probably just use DefaultEventLoopPolicy() instead of the type() thing, right? | 10:27 |
| juergbi | and we may also want to reset signal handlers. it might not be necessary but would probably be cleaner | 10:27 |
| benschubert | well, if the parent has set a policy, might be worth keeping it? But yes we could | 10:28 |
| juergbi | I'm also fine with type() if you prefer | 10:28 |
| juergbi | I hope that `del` (implicit or explicit) doesn't trigger any problematic code | 10:28 |
| benschubert | we'll see, worst case we can not del it | 10:28 |
| juergbi | if it does, it could make sense to explicitly keep that object (tree) around | 10:29 |
| juergbi | as we essentially just don't want it doing anything, but it's not like it would even free memory (used by parent) | 10:29 |
| benschubert | the memory would not be shared with the parent anymore | 10:29 |
| benschubert | Python uses ref-counting | 10:29 |
| benschubert | so first pass of the gc triggers a copy of all your program's pages :( | 10:30 |
| *** lachlan has joined #buildstream | 10:30 | |
| juergbi | ah, right | 10:31 |
| *** cs-shadow has joined #buildstream | 10:33 | |
| gitlab-br-bot | marge-bot123 closed issue #741 (Artifact cache configuration should have certs and keys expanded) on buildstream https://gitlab.com/BuildStream/buildstream/issues/741 | 10:35 |
| gitlab-br-bot | marge-bot123 merged MR !1696 (bschubert/expand-path-configs->master: _remote.py: Expand user in certificates paths) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1696 | 10:35 |
| *** lachlan has quit IRC | 10:40 | |
| *** lachlan has joined #buildstream | 10:48 | |
| *** lachlan has quit IRC | 11:04 | |
| *** lachlan has joined #buildstream | 11:06 | |
| *** lachlan has quit IRC | 11:09 | |
| *** lachlan has joined #buildstream | 11:13 | |
| *** lachlan_ has joined #buildstream | 11:20 | |
| *** lachlan has quit IRC | 11:23 | |
| *** lachlan_ has quit IRC | 11:23 | |
| *** lachlan has joined #buildstream | 11:30 | |
| benschubert | juergbi: I believe we can actually do even cleaner: | 11:30 |
| benschubert | ``` | 11:30 |
| benschubert | def run(self): | 11:30 |
| benschubert | signal.set_wakeup_fd(-1) | 11:30 |
| benschubert | old_child_watcher = get_child_watcher() | 11:30 |
| benschubert | set_child_watcher(None) | 11:30 |
| benschubert | del old_child_watcher | 11:30 |
| benschubert | old_event_loop = get_event_loop() | 11:30 |
| benschubert | set_event_loop(None) | 11:30 |
| benschubert | del old_event_loop | 11:30 |
| benschubert | super().run() | 11:31 |
| *** lachlan_ has joined #buildstream | 11:31 | |
| benschubert | ``` | 11:31 |
| benschubert | What do you think? | 11:31 |
| juergbi | benschubert: wouldn't this create an event loop if there is none yet? | 11:32 |
| juergbi | (and then immediately destroy it again) | 11:32 |
| benschubert | gah, that's correct,s ame for the child watcher | 11:33 |
| benschubert | it woul din practice never be the case, but fair point | 11:33 |
| benschubert | let's do the thing with the policy, it does work | 11:33 |
| benschubert | juergbi: then let's go for: | 11:34 |
| benschubert | ``` | 11:34 |
| benschubert | def run(self): | 11:34 |
| benschubert | signal.set_wakeup_fd(-1) | 11:34 |
| benschubert | set_event_loop_policy(None) | 11:34 |
| benschubert | super().run() | 11:34 |
| benschubert | ``` | 11:34 |
| benschubert | ? That does seem to work | 11:34 |
| juergbi | lgtm | 11:35 |
| gitlab-br-bot | traveltissues opened issue #1199 (Follow-up from "skip tracking elements without trackable sources") on buildstream https://gitlab.com/BuildStream/buildstream/issues/1199 | 12:05 |
| traveltissues | any further comments on !1689? | 12:08 |
| gitlab-br-bot | MR !1689: skip tracking elements without trackable sources https://gitlab.com/BuildStream/buildstream/merge_requests/1689 | 12:08 |
| benschubert | traveltissues: I still don't like the 'is not True' :) | 12:09 |
| traveltissues | benschubert, i changed that line | 12:09 |
| traveltissues | `if not self.__tracking_scheduled` | 12:10 |
| benschubert | wut? I'm not seeing this on gitlab, let me double check | 12:10 |
| traveltissues | me neither | 12:10 |
| traveltissues | strange | 12:10 |
| traveltissues | oops, one second | 12:11 |
| benschubert | good I'm not crazy :D | 12:11 |
| tlater[m] | You might both be | 12:11 |
| benschubert | tlater[m]: Don't ruin my hopes 😢 | 12:12 |
| tlater[m] | ;p | 12:12 |
| traveltissues | thanks benschubert i've added those | 12:19 |
| cs-shadow | marge tells me she's broken on the inside - https://gitlab.com/BuildStream/buildstream/merge_requests/1697#note_242801039. Anyone able to take a look please? :) | 12:38 |
| tlater[m] | tpollard should make his GitLab bio "Bot psychotherapist" or something. | 12:39 |
| tpollard | :( | 12:44 |
| * cs-shadow proposes "bot whisperer" | 13:15 | |
| * tpollard rolls https://siliconvalleyjobtitlegenerator.tumblr.com/ | 13:16 | |
| *** santi has quit IRC | 13:45 | |
| gitlab-br-bot | BenjaminSchubert approved MR !1689 (traveltissues/1186-3->master: skip tracking elements without trackable sources) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1689 | 13:50 |
| *** santi has joined #buildstream | 13:59 | |
| coldtom | could someone merge !1694 please? if nobody has any objections of course | 14:08 |
| gitlab-br-bot | MR !1694: _sandboxbwrap.py: Create /dev/shm in the sandbox https://gitlab.com/BuildStream/buildstream/merge_requests/1694 | 14:08 |
| gitlab-br-bot | marge-bot123 closed issue #1186 (Sources advertising no prestage key and not needing previous sources should skip tracking) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1186 | 14:17 |
| gitlab-br-bot | marge-bot123 merged MR !1689 (traveltissues/1186-3->master: skip tracking elements without trackable sources) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1689 | 14:17 |
| * coldtom just remembered marge exists | 14:35 | |
| coldtom | two approvals is good to merge right? | 14:35 |
| tlater[m] | coldtom: Approvals aren't even necessary anymore, unless your branch is very controversial I'd just go ahead. | 14:36 |
| traveltissues | personally i treat approvals as necessary on my MRs | 14:40 |
| tlater[m] | traveltissues: So do I, but the policy of "at least two approvals by core members" isn't in place anymore. | 14:40 |
| traveltissues | ik, but at least one is good | 14:41 |
| tlater[m] | Also, GitLab likes removing approvals if you modify anything. | 14:41 |
| traveltissues | that's sensible to do | 14:41 |
| coldtom | s/modify/rebase and push/ | 14:41 |
| coldtom | but yes, still sensible | 14:41 |
| tlater[m] | Absolutely, but I don't think prodding two people to re-approve makes sense in the rebase and push case ;) | 14:42 |
| traveltissues | well that wasn't the question but yes | 14:42 |
| gitlab-br-bot | BenjaminSchubert closed issue #1144 (bst shell -b ${ELEMENT} throws assertion when source not present) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1144 | 14:53 |
| gitlab-br-bot | marge-bot123 closed issue #1197 (/dev/shm not available in sandbox (bwrap)) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1197 | 15:15 |
| gitlab-br-bot | marge-bot123 merged MR !1694 (coldtom/allow-dev-shm->master: _sandboxbwrap.py: Create /dev/shm in the sandbox) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1694 | 15:15 |
| gitlab-br-bot | juergbi opened MR !1699 (juerg/sandbox-reapi->master: Add SandboxREAPI class) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1699 | 15:53 |
| *** lachlan_ has quit IRC | 15:54 | |
| *** lachlan has quit IRC | 15:54 | |
| *** lachlan has joined #buildstream | 16:05 | |
| *** lachlan_ has joined #buildstream | 16:06 | |
| tpollard | benschubert: hi! did you see my comment re running the subprocess branch on WSL with the profile set? | 16:06 |
| benschubert | tpollard: it slipped through the cracks, do you have a link? | 16:06 |
| tpollard | yep, one mo | 16:07 |
| gitlab-br-bot | cs-shadow opened MR !1700 (chandan/refactor-default-target->master: tests/frontend: Refactor tests for default targets) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1700 | 16:07 |
| *** phildawson has quit IRC | 16:07 | |
| *** ikerperez has quit IRC | 16:07 | |
| *** qinusty has quit IRC | 16:07 | |
| *** adds68 has quit IRC | 16:07 | |
| tpollard | benschubert: https://gitlab.com/BuildStream/buildstream/merge_requests/1613#note_241860854 | 16:07 |
| benschubert | I'll have a look thanks | 16:07 |
| *** qinusty has joined #buildstream | 16:07 | |
| tpollard | benschubert: cheers! | 16:19 |
| juergbi | tlater[m]: have you seen casd segfaults in your tests (i.e., casd died with exit code -11)? | 16:21 |
| tlater[m] | juergbi: Yes | 16:22 |
| tlater[m] | Caused by many a thing, usually permission issues | 16:23 |
| tlater[m] | Oh, wait, segfaults? | 16:23 |
| * tlater[m] has only really seen grpc refusals | 16:23 | |
| juergbi | yes, I wouldn't expect segfaults due to permission issues | 16:23 |
| juergbi | ok, will try to reproduce it locally and investigate | 16:24 |
| * tlater[m] is currently trying to figure out why the integration tests' base element's source ends up empty | 16:25 | |
| *** ikerperez has joined #buildstream | 16:30 | |
| *** lachlan_ has quit IRC | 16:36 | |
| *** lachlan has quit IRC | 16:36 | |
| *** lachlan_ has joined #buildstream | 16:38 | |
| *** lachlan has joined #buildstream | 16:38 | |
| gitlab-br-bot | BenjaminSchubert approved MR !1699 (juerg/sandbox-reapi->master: Add SandboxREAPI class) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1699 | 16:39 |
| *** lachlan has quit IRC | 16:40 | |
| *** lachlan_ has quit IRC | 16:40 | |
| *** lachlan has joined #buildstream | 16:40 | |
| *** santi has quit IRC | 16:43 | |
| *** lachlan has quit IRC | 16:43 | |
| *** lachlan has joined #buildstream | 16:47 | |
| *** phil has joined #buildstream | 17:08 | |
| *** phildawson_ has quit IRC | 17:09 | |
| *** santi has joined #buildstream | 17:11 | |
| *** traveltissues has quit IRC | 17:25 | |
| *** toscalix has quit IRC | 17:33 | |
| gitlab-br-bot | BenjaminSchubert opened MR !1701 (bschubert/backport-1690->bst-1: app.py: Also catch SystemError with click.Abort) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1701 | 17:49 |
| gitlab-br-bot | marge-bot123 merged MR !1697 (chandan/no-command-general-elements->master: Indicate that compose & stack plugins do not run commands) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1697 | 17:56 |
| *** santi has quit IRC | 18:09 | |
| *** lachlan has quit IRC | 18:21 | |
| *** lachlan has joined #buildstream | 18:22 | |
| *** jonathanmaw has quit IRC | 18:29 | |
| *** lachlan has quit IRC | 18:31 | |
| *** lachlan has joined #buildstream | 18:34 | |
| gitlab-br-bot | marge-bot123 merged MR !1691 (bschubert/fix-children-termination->master: scheduler.py: Prevent the asyncio loop from leaking into subprocesses) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1691 | 18:36 |
| *** lachlan has quit IRC | 18:39 | |
| *** tiagogomes has quit IRC | 18:46 | |
| gitlab-br-bot | BenjaminSchubert opened (was WIP) MR !1692 (bschubert/graceful-children-sigterm->master: Gracefully shutdown children on termination) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1692 | 18:48 |
| *** lachlan has joined #buildstream | 18:53 | |
| benschubert | https://gitlab.com/BuildStream/buildstream/merge_requests/1692 is ready for a final review :) | 18:54 |
| gitlab-br-bot | cs-shadow opened MR !1702 (chandan/fix-implicit-string-concat->master: tests/frontend/completions: Fix implicit string concatenation) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1702 | 18:54 |
| gitlab-br-bot | marge-bot123 merged MR !1699 (juerg/sandbox-reapi->master: Add SandboxREAPI class) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1699 | 19:04 |
| *** narispo has joined #buildstream | 20:16 | |
| *** lachlan has quit IRC | 20:24 | |
| *** lachlan has joined #buildstream | 20:32 | |
| *** lachlan has quit IRC | 20:48 | |
| *** lachlan has joined #buildstream | 20:57 | |
| *** lachlan has quit IRC | 21:00 | |
| *** lachlan has joined #buildstream | 21:10 | |
| *** lachlan has quit IRC | 21:11 | |
| *** narispo has quit IRC | 22:23 | |
| *** narispo has joined #buildstream | 22:23 | |
| *** cs-shadow has quit IRC | 22:23 | |
| gitlab-br-bot | cs-shadow opened issue #1200 (`bst init` does not respect `--no-colors`) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1200 | 22:51 |
| gitlab-br-bot | cs-shadow opened MR !1704 (chandan/fix-no-color->master: _frontend: Simplify color handling) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1704 | 23:26 |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!