| *** nimish has joined #buildstream | 01:42 | |
| *** mohan43u has quit IRC | 01:54 | |
| *** nimish has quit IRC | 03:53 | |
| *** nimish has joined #buildstream | 04:02 | |
| *** nimish has quit IRC | 06:13 | |
| Kinnison | Is there any chance we can get https://gitlab.com/BuildStream/buildstream/merge_requests/1116 resolved today? I am fed up with carrying a delta to setup.cfg in a git stash | 08:18 |
|---|---|---|
| *** cs-shadow has joined #buildstream | 08:54 | |
| gitlab-br-bot | BenjaminSchubert approved MR !1116 (doraskayo/exclude-eggs-from-linting->master: setup.cfg: exclude .eggs/** and build/** from pycodestyle linting) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1116 | 09:30 |
| *** tpollard has joined #buildstream | 09:51 | |
| cs-shadow | jjardon: benschubert: hi :) thoughts on https://gitlab.com/BuildStream/buildstream-docker-images/merge_requests/108 ? | 09:52 |
| jjardon | cs-shadow: nice! +1 | 09:56 |
| benschubert | cs-shadow: just not sure about "supported tags", don't know what support would entail here :) But otherwise, thansk a lot! | 09:56 |
| *** raoul has joined #buildstream | 09:57 | |
| jjardon | cs-shadow: not sure you noticed but master of that repo was failing yesterday | 09:59 |
| cs-shadow | jjardon: yes, that was because i hadn't created the buildstream/buildstream image on Docker Hub. I wanted to see it fail once to verify that i hadn't messed up the moving tags logic. Should be good now :) | 10:00 |
| cs-shadow | benschubert: i agree, that was inspired by "official" images but you are right. How about `s/supported tags/list of tags/` ? | 10:01 |
| benschubert | "published tags"? otherwise this seems good too! | 10:01 |
| jjardon | cs-shadow: coolio | 10:02 |
| cs-shadow | cool, "published" wins. thanks! | 10:02 |
| cs-shadow | Updated https://hub.docker.com/r/buildstream/buildstream | 10:07 |
| laurence | adds68, so, the marge-bot: just to check what it does - when an MR is approved and set to merge, it'll put it into the correct 'queue' and re-base so the user doesn't have to | 10:08 |
| laurence | is there any more to it? | 10:08 |
| Kinnison | I think you assign the MR to marge | 10:09 |
| Kinnison | and *it* deals with it | 10:09 |
| laurence | also is there anywhere else this is captured, btw? not sure where marge bot came from originally | 10:09 |
| laurence | yeah i'm aware of that, but you have to review it first before assigning | 10:09 |
| laurence | and approve, I think | 10:09 |
| adds68 | laurence, yea you need the standard number of approvals and then to merge it you just assign it to marge | 10:10 |
| adds68 | she then rebases/merges etc | 10:10 |
| laurence | great, so if you've got a few people trying to merge at once it handles that for you, very useful | 10:10 |
| laurence | adds68, just wanted to check if there was anything else it did to be aware of | 10:11 |
| adds68 | laurence, yea correct, i'm not sure if it has any other features | 10:13 |
| *** jonathanmaw has joined #buildstream | 10:16 | |
| laurence | great, thanks, I'd like to get it into BuildStream (very soon, I hope) | 10:19 |
| *** tristan has joined #buildstream | 10:20 | |
| *** lachlan has joined #buildstream | 10:36 | |
| *** alatiera has joined #buildstream | 10:44 | |
| gitlab-br-bot | tpollard opened issue #913 (Extracting buildtree/specific subdirs from the cache should be done directly) on buildstream https://gitlab.com/BuildStream/buildstream/issues/913 | 11:00 |
| *** ChanServ sets mode: +o tristan | 11:02 | |
| *** lachlan has quit IRC | 11:04 | |
| Kinnison | Can tox run an arbitrary script inside its virtual environment? | 11:14 |
| Kinnison | I want to run tests/cachekey/update.py but I can't because pytest et al are managed by tox | 11:14 |
| tpollard | you can run specific tests yes | 11:15 |
| Kinnison | it's not a test | 11:15 |
| Kinnison | it's just a script which ends up importing pytest | 11:15 |
| tpollard | ah | 11:16 |
| phildawson | Kinnison, To do so you'd have to add the script you want to run to the 'commands' in tox.ini | 11:16 |
| Kinnison | aha, ta | 11:16 |
| phildawson | As far as I'm aware, there's no way of doing so by passing arguments to tox, but I'm happy to be corrected on that | 11:16 |
| * Kinnison tries creating a cachekey-update env | 11:18 | |
| benschubert | otherwise: source .tox/{env-name}/bin/activate and you're in the env :D | 11:18 |
| benschubert | and afterwards you can do whatever | 11:19 |
| *** lachlan has joined #buildstream | 11:20 | |
| Kinnison | aha so they are venv-alike? | 11:21 |
| benschubert | they are venvs | 11:22 |
| Kinnison | nice | 11:22 |
| benschubert | tox is basically creating one venv per "env" you declare in tox | 11:22 |
| benschubert | and that's it | 11:22 |
| Kinnison | the env thing turned out easiest, thanks benschubert | 11:24 |
| * phildawson wishes he'd thought of that earlier. Thanks benschubert :) | 11:28 | |
| SotK | phildawson: you can pass args by putting `{posargs}` in the command | 11:30 |
| SotK | so you can do something like http://git.openstack.org/cgit/openstack-infra/storyboard/tree/tox.ini#n31 to just run whatever you pass inside the venv managed by tox | 11:31 |
| phildawson | SotK, You can, but the way buildstream's tox.ini is set up all those are passed straight to pytest | 11:32 |
| phildawson | Which makes it somewhat inconvenient if you just want to run some arbitrary script :) | 11:32 |
| SotK | the tox.ini I linked does similar for the default env, but has a `venv` environment in which the posargs are the entire command | 11:34 |
| SotK | so `tox -e venv -- my-command-here` works | 11:34 |
| phildawson | SotK, Neat. I might add that to ours :) | 11:36 |
| gitlab-br-bot | aevri opened issue #914 (checkout --tar: mtime=0 is problematic for some software) on buildstream https://gitlab.com/BuildStream/buildstream/issues/914 | 11:40 |
| gitlab-br-bot | aevri opened MR !1149 (aevri/mtime1->master: storage.Directory.export_to_tar: default mtime=1) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1149 | 11:43 |
| adds68 | laurence, the designer finally sent BuildStream sources back | 11:43 |
| adds68 | laurence, should i upload them to that issue, so we can close it and someone can set the logo on the project? | 11:43 |
| juergbi | valentind: I assume you're planning to comment again on !1140 and I will hold off merging for a bit. let me know if/when you're happy with it | 11:45 |
| gitlab-br-bot | MR !1140: Do not resolve or mangle symlinks during staging https://gitlab.com/BuildStream/buildstream/merge_requests/1140 | 11:45 |
| laurence | adds68, for BuildGrid I just uploaded the logo sources to the gitlab wiki, for us maybe the GNOME wiki is better (just wanted them somewhere so others could always access) | 11:49 |
| laurence | adds68, for the question of BuildStream versus Buildstream - has the designer amended it on the sources? | 11:50 |
| gitlab-br-bot | BenjaminSchubert opened MR !1150 (bschubert/dont-keep-metasource->master: Don't keep MetaSource around in Source) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1150 | 12:02 |
| tristan | jjardon, I think we should wait for Kinnison's emergency fix for cache key stability to drop before rolling out a 1.2.x | 12:08 |
| tristan | jjardon, i.e. https://mail.gnome.org/archives/buildstream-list/2019-February/msg00035.html, it is a cache key breaking change; but seems to be very important (i.e. also a cache key stability *fixing* change) | 12:09 |
| juergbi | oh, breaking cache key in 1.2? | 12:09 |
| tristan | Well | 12:09 |
| tristan | If they are broken ? | 12:09 |
| juergbi | well, it's more like Python doesn't guarantee that it's stable | 12:09 |
| tristan | I mean... they are broken right ? | 12:09 |
| juergbi | so it might break with future Python version | 12:09 |
| juergbi | as we test cache keys in CI with Python 3.5-3.7, it should be fine right now, or not? | 12:09 |
| juergbi | (as long as we also don't backport optimizations that break pickle/cache key) | 12:10 |
| jmac | Do you actually test cache keys from Python 3.5 against keys from 3.7? | 12:10 |
| tristan | This is a good question | 12:10 |
| juergbi | yes, CI tests that | 12:10 |
| jmac | Nice | 12:10 |
| juergbi | although we can't guarantee that we test all corner cases | 12:10 |
| juergbi | Kinnison: any opinions on the above? | 12:11 |
| tristan | Yeah, I don't think I fully grasp why pickles differ depending on 'identity' (address of module in memory ?) | 12:12 |
| Kinnison | It's that pickles try not to serialise the same object more than once | 12:39 |
| Kinnison | if you have two strings which happen to have come by the value "foo" independently they get serialised independently | 12:39 |
| Kinnison | but if yo uhave two strings which have the same value "foo" then they get serialised once and recalled later | 12:39 |
| Kinnison | This makes relying on pickle dangerous in the face of seemingly innocuous optimisations elsewhere in the code | 12:40 |
| Kinnison | I think, if we're already changing 1.2's cache keys, then it makes sense to include the JSON change too | 12:44 |
| Kinnison | but if we're not, then there's no point breaking 1.2's users' caches | 12:44 |
| tristan | Kinnison, in that case lets not | 12:44 |
| * Kinnison nods | 12:44 | |
| tristan | Kinnison, policy is to not break keys in a release - what I don't understand is; is it potentially broken ? | 12:45 |
| Kinnison | tristan: So long as the cache keys tests are in place, it's guaranteed not broken for the set of python versions we CI | 12:45 |
| tristan | The description looks like we might get the wrong cache key if two elements are supposed to have the same key ? | 12:45 |
| tristan | Depending on what strings have been seen in the program lifetime ? | 12:46 |
| Kinnison | tristan: essentially the same cache key dictionary *value* could result in different pickles, dependent on whether or not identical *value* strings happen to be the same PyObject under the hood | 12:46 |
| Kinnison | tristan: the difference between foostring == barstring, and foostring is barstring | 12:46 |
| Kinnison | IYSWIM | 12:46 |
| tristan | Hmm | 12:47 |
| tristan | That seems like it is probably happening, but it is happening in a consistent way and as such things are not broken ? | 12:48 |
| Kinnison | Currently yes | 12:48 |
| Kinnison | But, like I said, an unrelated optimisation in _variables.py exposed this | 12:48 |
| tristan | I.e. a cached cache key string will be reused various times (by virtue of collecting the dependency cache keys and including those) | 12:48 |
| *** raoul has quit IRC | 12:49 | |
| jmac | I think so. I wouldn't have expected Python 3.5 and 3.7 to produce consistent pickles | 12:49 |
| jmac | Standard advice from the Python community is never to have a pickled object last longer than the Python process which created it | 12:49 |
| tristan | I see, so only code changes appear to have an effect on this, and currently we have consistent keys across python versions | 12:49 |
| Kinnison | yes | 12:49 |
| Kinnison | afaict, run-to-run we're consistent | 12:49 |
| Kinnison | It's just in the face of code changes to the code which provides the data which may end up in the cache key | 12:50 |
| jmac | What happens if you alter the YAML in one .bst file which means it duplicates a string value in another .bst file? | 12:50 |
| Kinnison | We're fine because we generate fresh strings - there's no interning in ruamel | 12:51 |
| Kinnison | The issue came up because I added an optimisation in my Variables rework to go "oh this is constant, just return it" | 12:52 |
| Kinnison | rather than always generating a fresh string by means of re.subst | 12:52 |
| tristan | I suppose we also have to change _yamlcache.py | 12:52 |
| Kinnison | I think the yaml cache is play | 12:53 |
| Kinnison | okay | 12:53 |
| * Kinnison can type, honest | 12:53 | |
| Kinnison | The yaml cache is really only for run-to-run of the same code | 12:53 |
| Kinnison | And future python commits to always being able to read old pickle | 12:53 |
| gitlab-br-bot | danielsilverstone-ct opened MR !1151 (danielsilverstone-ct/json-cache-key->master: Update cache keys to use JSON) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1151 | 12:55 |
| tristan | Ok well anyway, this is all a lot less alarming than I thought :) | 12:56 |
| Kinnison | I have put an MR up, we'll see if the pipeline is okay or if I've missed anything, my local test runs were okay, but I only have 3.5 on my laptop | 12:56 |
| Kinnison | tristan: yeah it's not terrifyingly bad, just irritating in the face of optimisation work I'm trying to do | 12:56 |
| Kinnison | if Python always interned strings, this'd have not come to my attention | 12:56 |
| Kinnison | Ditto if pickle were less efficient :D | 12:57 |
| Kinnison | As it is, I'm proposing to use `ujson` which ends up a few seconds faster than the old way on my test set anyway | 12:57 |
| Kinnison | That, plus the _variables.py rewrite it unblocks, is about a 30s improvement on runtime for `bst show debian-stack.bst` on my laptop | 12:58 |
| Kinnison | which is ca. 10% | 12:58 |
| tristan | Should we squash the commits on your branch for the sake of bisectability ? | 12:58 |
| tristan | Kinnison, that sounds great re _variables.py :) | 12:59 |
| Kinnison | tristan: For a long time I've assumed we can only bisect between merge commits | 13:00 |
| Kinnison | tristan: otherwise, given the requirement on merges being fastforwardable, I'd have assumed we'd be ffwding | 13:00 |
| Kinnison | I don't mind squashing them if you want | 13:00 |
| tristan | We try to keep every commit in a series passing it's tests indeed | 13:01 |
| Kinnison | I'll squash them if that's the ideal :D | 13:01 |
| tristan | Yeah, I also dislike squashing but I think this is a good reason :) | 13:02 |
| Kinnison | squashed | 13:02 |
| Kinnison | benschubert: I think https://gitlab.com/BuildStream/buildstream/merge_requests/1150 looks good :+1: | 13:02 |
| * Kinnison likes that it removes a FIXME | 13:03 | |
| *** alatiera has quit IRC | 13:03 | |
| lachlan | Hi, I'm getting 404 from https://registry.gitlab.com/buildstream/buildstream-docker-images - can others confirm please. | 13:08 |
| tristan | same here | 13:09 |
| *** alatiera has joined #buildstream | 13:10 | |
| *** tristan changes topic to "BuildStream 1.2.4 is out ! | https://gitlab.com/BuildStream/buildstream | Docs: https://docs.buildstream.build/ | IRC logs: https://irclogs.baserock.org/buildstream | Mailing List: https://mail.gnome.org/mailman/listinfo/buildstream-list | Roadmap: https://wiki.gnome.org/Projects/BuildStream/Roadmaps" | 13:15 | |
| *** raoul has joined #buildstream | 13:17 | |
| cs-shadow | lachlan: that’s not the url for the web UI. That only works with Docker commands (without https) | 13:18 |
| cs-shadow | For browsing, please use https://gitlab.com/BuildStream/buildstream-docker-images/container_registry | 13:18 |
| Kinnison | tristan: congrats on the release | 13:19 |
| tristan | tpollard, w00t, last minute proposal for BuildStream talk at FOSSASIA ... (drumroll)... accepted ! | 13:19 |
| Kinnison | yay | 13:20 |
| tpollard | tristan: awesome! I sent in my request to attend today :) | 13:21 |
| Kinnison | tristan: 축하해 | 13:21 |
| tristan | Kinnison, Thanks ! mostly valentind was the powerhouse behind 1.2.4 :) | 13:21 |
| Kinnison | valentind: félicitations | 13:22 |
| valentind | What what? | 13:22 |
| valentind | New 1.2 release? | 13:22 |
| tristan | Yeah | 13:22 |
| valentind | Cool! | 13:22 |
| tristan | valentind, it's finally out, thanks jjardon for nagging heh :) | 13:23 |
| jjardon | tristan: \o/ tvm | 13:23 |
| * jjardon fires CI on everything | 13:23 | |
| valentind | There are key changes in filter elements, so few things will build. | 13:24 |
| tristan | Kinnison, I just learned 축하, and now the Korean version of "Happy birthday" makes sense :) | 13:24 |
| * cs-shadow is on mobile at the moment. Will publish 1.2.4 on PyPI and Docker Hub once a computer is nearby | 13:26 | |
| jonathanmaw | juergbi: I think I might be misunderstanding your comment at https://gitlab.com/BuildStream/buildstream/issues/911#note_140969502. Is the fork callback the function passed as "target" in the Process' constructor? I have doubts because I tried adding an artificial delay to Elementjob.child_process and didn't see an increase in time spent in posix.fork | 13:28 |
| gitlab-br-bot | tpollard approved MR !1148 (chandan/dot-graph->master: contrib/bst-graph: Add script to print graph in DOT format) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1148 | 13:28 |
| Kinnison | tristan: :D | 13:29 |
| adds68 | laurence, ok i will try and upload them | 13:42 |
| Kinnison | tristan: Now that it's squshed, are you okay with https://gitlab.com/BuildStream/buildstream/merge_requests/1151 ? | 13:42 |
| adds68 | laurence, and yea he has | 13:42 |
| juergbi | jonathanmaw: no, I mean pthread_atfork() and the Python equivalent register_at_fork or similar | 13:42 |
| tristan | Kinnison, Yes ! | 13:42 |
| jonathanmaw | juergbi: oh, ta | 13:42 |
| juergbi | it's just a guess, though | 13:42 |
| * tristan hits merge when ready button | 13:43 | |
| tristan | Actually I was counting on a pipeline running on master to update the badges in the docs | 13:43 |
| Kinnison | heh | 13:43 |
| Kinnison | Looks like the pipelines left on 1151 are just at the cache stage so should be done soon anyway | 13:44 |
| adds68 | laurence, seems i don't have access to upload things | 13:44 |
| tristan | Kinnison, Yeah, I don't know that those docs were built from a repo state which contained the new tag, but it will certainly with your branch landing | 13:45 |
| * Kinnison nods | 13:45 | |
| Kinnison | Hmm cache/: found 677332 matching files | 13:45 |
| Kinnison | I wonder if we're caching too much | 13:45 |
| Kinnison | (gitlab-ci) | 13:45 |
| tristan | Oh no, not that again | 13:45 |
| Kinnison | Something we've seen before? | 13:46 |
| tristan | Kinnison, FYI, common issue with gitlab is that the caches are like, shared | 13:46 |
| tristan | So... when someone runs CI on their branch and that branch adds a ton of data to the cache... it doesnt go away | 13:47 |
| Kinnison | aah | 13:47 |
| tristan | I had a branch which force removes stuff which I manually run a bunch of pipelines on | 13:47 |
| tristan | yeah, it must be happening, CI is taking absurdly long :-/ | 13:48 |
| laurence | adds68, ok, send to me and I'll put them up, cheers | 13:52 |
| adds68 | laurence, done | 13:53 |
| tpollard | adds68: make sure you have stickers for tristan at FOSSASIA ;) | 13:53 |
| adds68 | haha tristan get some ordered from Sticker Mule | 13:54 |
| adds68 | $20 discount for FOSS projects | 13:54 |
| tristan | Oh ! ok ok | 13:55 |
| Kinnison | tristan: Can we clear the caches? Also can we make it so that the caches are only retained from runs on master? | 13:56 |
| tristan | I think we can clear them, I don't know if we can do the latter though | 13:56 |
| tristan | Maybe we should clear them on a timer, like every sunday | 13:56 |
| gitlab-br-bot | tristanvb merged MR !1151 (danielsilverstone-ct/json-cache-key->master: Update cache keys to use JSON) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1151 | 13:58 |
| *** phildawson has quit IRC | 13:59 | |
| *** phildawson has joined #buildstream | 13:59 | |
| tristan | jjardon, Do you know if there is a way to zap the caches in the gitlab UI ? I can't find one | 14:02 |
| *** phildawson has quit IRC | 14:05 | |
| juergbi | tristan: there is a 'Clear Runner Caches' on https://gitlab.com/BuildStream/buildstream/pipelines | 14:12 |
| tristan | Ah on the pipelines page ? | 14:13 |
| * tristan was fooling around in the settings | 14:13 | |
| tristan | Kinnison, it's a bit late for me to pick up pieces if things break... but I'll try the button tomorrow | 14:15 |
| benschubert | Could someone have a look at https://gitlab.com/BuildStream/buildstream/merge_requests/1150 ? | 14:27 |
| tristan | benschubert, Nice ! | 14:28 |
| tristan | Assuming the tests pass, that should free up some memory :) | 14:28 |
| gitlab-br-bot | LaurenceUrhegyi opened issue #915 (Artifact As A Proto: Capabilities Service) on buildstream https://gitlab.com/BuildStream/buildstream/issues/915 | 14:29 |
| gitlab-br-bot | danielsilverstone-ct opened MR !1152 (danielsilverstone-ct/variables-rework->master: Variables: Rework how expansion strings work) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1152 | 14:29 |
| *** phildawson has joined #buildstream | 14:30 | |
| gitlab-br-bot | aevri opened (was WIP) MR !1149 (aevri/mtime1->master: storage.Directory.export_to_tar: default mtime=utils._magic_timestamp) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1149 | 14:47 |
| *** tristan has quit IRC | 14:53 | |
| *** lachlan has quit IRC | 14:56 | |
| juergbi | alatiera: maybe I misunderstood your question. my branch fixes build with bst master but I wasn't proposing to switch CI to bst master. not sure whether jjardon wants to do that | 15:03 |
| laurence | would anyone have any objections to the merge-bo | 15:04 |
| laurence | gah | 15:04 |
| laurence | merge-bot? I can't see why, since it will make lives easier, but you never know, so thought i'd ask | 15:04 |
| gitlab-br-bot | tristanvb merged MR !1150 (bschubert/dont-keep-metasource->master: Don't keep MetaSource around in Source) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1150 | 15:04 |
| alatiera | juergbi: I think jjardon wants the tag of the CI to be bumped to the new bst 1.2 release, my question was towards that not about your changes :P | 15:04 |
| juergbi | ok, sorry for the confusion | 15:04 |
| alatiera | its fine :) | 15:05 |
| alatiera | juergbi: thanks for the work again :) | 15:05 |
| juergbi | yw | 15:05 |
| * laurence wonders who has the image for the mascot - tlater[m], any idea? | 15:10 | |
| tlater[m] | laurence: Sam bought that from somewhere | 15:11 |
| tlater[m] | (Thursfield) | 15:11 |
| tpollard | laurence: https://gitlab.com/BuildStream/nosoftware/communication/blob/master/logo_icon/buildstream-beaver.jpg | 15:11 |
| laurence | ah, it's in nosoftware | 15:12 |
| laurence | thanks tlater[m] | 15:12 |
| tlater[m] | Looks like we have a good copy then :) | 15:12 |
| *** lachlan has joined #buildstream | 15:13 | |
| *** nimish has joined #buildstream | 15:23 | |
| *** rdale has quit IRC | 15:25 | |
| jjardon | Kinnison: caches can be configured per branch/job/wherever; the only thing needed is to configure correctly the cache key in the .cache: section in the gitlab-ci.yml | 15:31 |
| jjardon | oh, no tristan | 15:31 |
| * jjardon not sure what he means by " there is a way to zap the caches in the gitlab UI" | 15:32 | |
| *** rdale has joined #buildstream | 15:40 | |
| *** lachlan has quit IRC | 15:42 | |
| *** tristan has joined #buildstream | 15:42 | |
| *** lachlan has joined #buildstream | 15:49 | |
| *** rdale has quit IRC | 15:59 | |
| * paulsherwood reads Kinnison's email about cach-key stability and sighs | 16:04 | |
| paulsherwood | 1 - isn't this tested in CI? | 16:04 |
| paulsherwood | 2 - why was pickle used in the first place? | 16:05 |
| Kinnison | paulsherwood: 1. yes, but CI never hit the issue, I did 2. as Tristan said - convenience at the time. | 16:05 |
| Kinnison | paulsherwood: and 3. We've sorted that now | 16:05 |
| paulsherwood | in 'sorting' it have we assured backwards compatibility? | 16:06 |
| Kinnison | We are now using JSON as the serialisation form for the cache key dictionary which means we won't hit this kind of confusion in future | 16:06 |
| Kinnison | This change has nothing whatsoever to do with being able to find artifacts from previous algorithms | 16:06 |
| paulsherwood | you didn't answer my question | 16:06 |
| Kinnison | let me answer with a question then -- "backwards compatibility" in what sense? | 16:07 |
| paulsherwood | can the new implementation generate the same (assumed correct) hashes as the previous one | 16:07 |
| Kinnison | No, because the previous serialisation format was pickle and the new serialisation format is json | 16:08 |
| gitlab-br-bot | juergbi merged MR !1140 (juerg/symlinks2->master: Do not resolve or mangle symlinks during staging) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1140 | 16:08 |
| Kinnison | And the point of changing it was that the old serialisation format resulted in cache keys changing with unrelated optimisations elsewhere in the code, due to pickle being a little too clever for our own good | 16:09 |
| paulsherwood | Kinnison: https://gitlab.com/baserock/ybd/blob/master/ybd/cache.py#L67 | 16:09 |
| Kinnison | Yes, that looks similar to what we've merged into buildstream today | 16:09 |
| paulsherwood | yes. that predates the existence of bst. just sayin' :/ | 16:10 |
| Kinnison | Not a terribly constructive comment, but true nonetheless | 16:10 |
| paulsherwood | after so many years of being told by folks that i don't understand the problem, i do struggle to be 'constructive' at times yes | 16:11 |
| paulsherwood | i'll get my coat :) | 16:11 |
| *** paulsherwood has left #buildstream | 16:19 | |
| gitlab-br-bot | juergbi closed issue #390 ("FAILURE Failed to link ... No such file or directory" in a compose element) on buildstream https://gitlab.com/BuildStream/buildstream/issues/390 | 16:23 |
| gitlab-br-bot | juergbi closed issue #606 (Mangled symlinks when importing freedesktop BasePlatform) on buildstream https://gitlab.com/BuildStream/buildstream/issues/606 | 16:24 |
| gitlab-br-bot | juergbi closed MR !1023 (jmac/830-stop-resolving-symlinks->master: Stop resolving symlinks in _relative_symlink_target) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/1023 | 16:28 |
| *** rdale has joined #buildstream | 16:28 | |
| gitlab-br-bot | juergbi closed issue #830 (_relative_symlink_target depends on the host filing system) on buildstream https://gitlab.com/BuildStream/buildstream/issues/830 | 16:29 |
| *** lachlan has quit IRC | 16:38 | |
| gitlab-br-bot | juergbi closed MR !862 (valentindavid/link_files_sort_resolved->master: Resolve paths before ordering them in link_files/copy_files) on buildstream https://gitlab.com/BuildStream/buildstream/merge_requests/862 | 16:38 |
| gitlab-br-bot | juergbi closed issue #647 (BuildStream misorders elements in subdirectories of symlinks when link_files called in compose elements) on buildstream https://gitlab.com/BuildStream/buildstream/issues/647 | 16:42 |
| *** rdale has quit IRC | 16:44 | |
| *** cs-shadow has quit IRC | 16:54 | |
| *** rdale has joined #buildstream | 17:02 | |
| *** lachlan has joined #buildstream | 17:03 | |
| *** alatiera has quit IRC | 17:04 | |
| *** alatiera has joined #buildstream | 17:10 | |
| *** jonathanmaw has quit IRC | 18:18 | |
| *** lachlan has quit IRC | 18:18 | |
| *** raoul has quit IRC | 18:32 | |
| *** slaf has quit IRC | 18:40 | |
| *** slaf has joined #buildstream | 18:47 | |
| *** nimish_ has joined #buildstream | 18:51 | |
| *** nimish has quit IRC | 18:51 | |
| *** nimish_ is now known as nimish | 18:51 | |
| *** nimish has quit IRC | 18:52 | |
| *** nimish has joined #buildstream | 18:53 | |
| *** alatiera has quit IRC | 19:00 | |
| *** alatiera has joined #buildstream | 19:03 | |
| *** nimish has quit IRC | 19:47 | |
| *** nimish has joined #buildstream | 20:11 | |
| *** benschubert has quit IRC | 20:36 | |
| *** alatiera has quit IRC | 21:26 | |
| *** tristan has quit IRC | 21:31 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!
