| *** traveltissues has joined #buildstream | 08:37 | |
| *** phildawson has joined #buildstream | 08:40 | |
| *** traveltissues has quit IRC | 09:11 | |
| *** traveltissues has joined #buildstream | 09:12 | |
| gitlab-br-bot | juergbi opened MR !1807 (juerg/buildbox-run-bind-mount->master: _sandboxbuildboxrun.py: Support host-files) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1807 | 09:28 |
|---|---|---|
| *** tme5 has joined #buildstream | 09:32 | |
| gitlab-br-bot | tmewett opened MR !1808 (tmewett/git-source->master: Improvements to _GitSourceBase and _GitMirror) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1808 | 09:47 |
| *** phildawson has quit IRC | 10:02 | |
| *** phildawson has joined #buildstream | 10:03 | |
| gitlab-br-bot | BenjaminSchubert approved MR !1807 (juerg/buildbox-run-bind-mount->master: _sandboxbuildboxrun.py: Support host-files) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1807 | 10:05 |
| *** jonathanmaw has joined #buildstream | 10:17 | |
| tme5 | benschubert, thanks for your comments, I've replied | 10:46 |
| coldtom | hi, if anyone gets time could i get a review (and hopefully a merge) on !1806 please? | 10:48 |
| gitlab-br-bot | MR !1806: Stop plugin tests using private API https://gitlab.com/BuildStream/buildstream/merge_requests/1806 | 10:48 |
| benschubert | tme5: do you know a vway we could avoid doing a 'describe' instead of a 'rev-parse' there? describe is 20 times slower at least, and becomes worse with bigger repos :/ | 10:53 |
| benschubert | if not, that's fine we can keep it like that until it becomes a more important problem, we are still sub second with 10k commits | 10:54 |
| tme5 | oh really? i didn't know that | 10:55 |
| benschubert | tme5: for the workspace case, before git 2.0, you would not be able to push anything until you would do 'git pull --depths=${bigger_than_your_repo_history}' which can be unsettling for people not knowing git that well. I don't know whether we should take that in consideration there or not. git < 2.0 means all of RHEL 7 (would need to check for RHEL8) | 10:56 |
| tme5 | it's just with ref-format: 'git-describe' describe has to be called anyway, so i thought this was an optimisation on the whole | 10:57 |
| benschubert | tme5: on the other hand it's only in 'track', which is ok-ish. But it's main process, so if you end up cloning the linux kernel, systemd and such, might be a hit | 10:58 |
| benschubert | Might be good testing on a big (public) repo :) | 10:58 |
| benschubert | tme5: on a big-ish repo (a few hundred megs), on my laptop under WSL: 'git rev-parse' takes 0.016s, 'git describe' takes 1.6s | 10:59 |
| tme5 | but this is still a problem though, right? when ref-format is changed? | 11:02 |
| benschubert | it's definitely not idea | 11:03 |
| benschubert | *ideal | 11:03 |
| benschubert | but not such of a big deal, so we could get it in like that and potimize later. But if we can know which one we need before running the command, I'd definitely try to optimize :) | 11:03 |
| tme5 | do you have any idea how long it takes to start and finish a subprocess call from python? | 11:05 |
| benschubert | based on a quick check doing: import time;import subprocess; def t(): s = time.time(); subprocess.check_call(["echo"]), print(time.time() - s), it's roughly 0.002s per call | 11:08 |
| benschubert | (not a real benchmark, really just a hack) | 11:08 |
| tme5 | faster than for some reason i thought it would be | 11:09 |
| benschubert | python is not that bad :) | 11:11 |
| tpollard | haha | 11:12 |
| tme5 | benschubert, w.r.t workspaces, i'm not sure I understand, are you saying if the workspace is opened without the source ever having done a full fetch, you will be unable to immediately push on old versions of Git? | 11:17 |
| benschubert | tme5: correct, and that might confuse most people :) I don't think we should cater for that, but might be worth checking with others before putting this change in, since it is a breaking change :) | 11:22 |
| *** cs-shadow has joined #buildstream | 11:31 | |
| tme5 | it makes me wonder whether tracking will work either when a full fetch hasn't happened | 11:40 |
| tme5 | i could just full-fetch before tracking and opening workspaces | 11:41 |
| tme5 | but then i wonder whether the extra complexity is worth it, for optimising such a narrow case | 11:42 |
| tme5 | actually i suppose it's still very useful if you're just using a buildstream project and not developing it, since you'll never track or open workspaces | 11:48 |
| benschubert | Or at least you won't be doing it for most elements :) | 11:49 |
| *** santi has joined #buildstream | 11:51 | |
| gitlab-br-bot | tpollard closed issue #1151 (`bst shell --use-buildtree ask` uses `--pull` inconsistently) on buildstream https://gitlab.com/BuildStream/buildstream/issues/1151 | 12:11 |
| gitlab-br-bot | tpollard merged MR !1767 (tpollard/shellbuildtree->master: _frontend/cli.py: Make show() --use-buildtree respect pull semantics) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1767 | 12:11 |
| *** phildawson has quit IRC | 12:12 | |
| benschubert | tme5: for the git version and problems with git <2.0, let's not care about it and fix it when someone complains otherwise? Might be simpler and would enable a much faster experience for users | 12:16 |
| tpollard | think about the transition of git/bzr/tar/autotools into experimental seems problematic to me | 12:41 |
| tpollard | adding to the mix that workspaces are technically source plugin now, but still core, but then require external input | 12:42 |
| tpollard | & source plugins being required in other core tests (such as cachekey for git/bzr), and what we want to classify doc tests as | 12:45 |
| tpollard | I like you idea of '@pytest.mark.plugins' benschubert, but I'm just worried everything could get a bit messy | 12:49 |
| coldtom | maybe core tests could minimise the dependency on external plugins, by either using ones to be left in core or creating some minimal plugin for testing | 12:51 |
| coldtom | i don't see a way around docs using external plugins if they want to be anything like real example projects | 12:52 |
| *** narispo has quit IRC | 12:55 | |
| tpollard | I agree on the docs test | 12:56 |
| tpollard | even if we went back to considering a dedicated source cache for the docs | 12:57 |
| *** phildawson has joined #buildstream | 13:00 | |
| *** phildawson has quit IRC | 13:24 | |
| benschubert | tpollard: For workspace tests, we could use 'local' sources. For cache-key, we could use them with plugins installed | 13:40 |
| benschubert | tpollard: any specific concerns around the 'mark.plugins' getting messy? | 13:40 |
| *** phildawson has joined #buildstream | 13:40 | |
| *** phildawson has quit IRC | 14:26 | |
| *** phildawson has joined #buildstream | 14:46 | |
| *** bochecha has joined #buildstream | 14:54 | |
| *** bochecha_ has joined #buildstream | 15:29 | |
| *** bochecha has quit IRC | 15:30 | |
| tpollard | benschubert: above what I mentioned above... I'm looking at moving pip source. Most of it is fine, but the pip element integration tests have a particular test that also requires the source plugin | 16:28 |
| *** phildawson has quit IRC | 16:28 | |
| tpollard | which would work with the 'mark.plugins' approach, but it's just another example of where they overlap | 16:32 |
| *** traveltissues has quit IRC | 16:34 | |
| tpollard | most of it's fine being the extraction side https://gitlab.com/BuildStream/bst-plugins-experimental/merge_requests/74 | 16:37 |
| coldtom | tpollard: should the pip element not be moved into bst-plugins-experimental anyway? | 16:44 |
| tpollard | yep | 16:44 |
| tpollard | was hoping that it could be done separately though | 16:45 |
| tpollard | but pip is a special case I guess | 16:46 |
| benschubert | tpollard: or move first the element? That should work :) | 17:10 |
| tpollard | yep! | 17:14 |
| benschubert | I'll send a ML post around the mark.plugins whenever I get time to look at the ML | 17:14 |
| tpollard | I'm glad pip is our only package manager in core | 17:15 |
| tpollard | benschubert: thanks! | 17:15 |
| *** santi has quit IRC | 17:25 | |
| *** tme5 has quit IRC | 17:28 | |
| *** santi has joined #buildstream | 17:47 | |
| *** jonathanmaw has quit IRC | 18:04 | |
| *** santi has quit IRC | 20:15 | |
| *** santi has joined #buildstream | 20:15 | |
| *** santi has quit IRC | 20:33 | |
| *** us has joined #buildstream | 20:34 | |
| *** us has left #buildstream | 20:34 | |
| *** santi has joined #buildstream | 23:41 | |
| *** santi has quit IRC | 23:45 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!