| *** jamielennox|away is now known as jamielennox | 00:06 | |
| prometheanfire | +requests>=2.1.0,!=2.4.0,<=2.2.1 | 01:21 |
|---|---|---|
| prometheanfire | LOL | 01:21 |
| *** takedakn has joined #openstack-stable | 01:59 | |
| *** takedakn1 has joined #openstack-stable | 02:16 | |
| *** takedakn has quit IRC | 02:17 | |
| *** takedakn1 has quit IRC | 02:29 | |
| *** takedakn has joined #openstack-stable | 02:29 | |
| *** takedakn has quit IRC | 02:37 | |
| tchaypo | prometheanfire: that makes perfect sense *nods* | 03:03 |
| prometheanfire | :P | 03:04 |
| prometheanfire | I commented on the review for this commit https://review.openstack.org/#/c/147451/ | 03:04 |
| *** mrunge has joined #openstack-stable | 06:13 | |
| *** mrunge has quit IRC | 06:16 | |
| *** mrunge has joined #openstack-stable | 06:17 | |
| *** mrunge has quit IRC | 06:17 | |
| *** rwsu has joined #openstack-stable | 06:34 | |
| *** e0ne has joined #openstack-stable | 06:41 | |
| *** e0ne has quit IRC | 06:42 | |
| *** mrunge has joined #openstack-stable | 06:46 | |
| ttx | prometheanfire: actually those are the requirements for stable branch testing. The <= is set to correspond to the current version when the cap is set | 06:47 |
| ttx | (or at least that's my understanding) | 06:47 |
| prometheanfire | ya, just read through it some more | 06:48 |
| ttx | Before we let the dependency range open, but that broke every week | 06:48 |
| prometheanfire | at this point I'm ignoring the caps that were put on | 06:48 |
| ttx | yes, that's basically what we did until now | 06:48 |
| prometheanfire | commented on https://review.openstack.org/#/c/147451/ | 06:48 |
| ttx | (consider that everything works even if they don't) | 06:48 |
| ttx | OK, I'll read that. Just catching up | 06:48 |
| prometheanfire | were any of the distros notified? | 06:48 |
| prometheanfire | also, those automated deps are hard to read | 06:49 |
| prometheanfire | some of them are funny | 06:49 |
| prometheanfire | Babel>=1.3,<=1.3 | 06:49 |
| prometheanfire | requests>=2.1.0,!=2.4.0,<=2.2.1 | 06:49 |
| ttx | That capping was discussed extensively at the last summit in Paris and in the ML at the beginning of the cycle. The thing is, we could not apply strict capping like this for our libraries since that would break them | 06:50 |
| prometheanfire | I'll likely be fine with this for kilo, but it was a surprise to me | 06:50 |
| ttx | so for our libs we force them to prperly use semver and do >=x.y.z <x.y+1.0 | 06:50 |
| prometheanfire | esp since we already removed some older versions that were pinned back | 06:51 |
| ttx | the problem being, other python libraries do not properly use semver | 06:51 |
| ttx | so for thoise we strictly cap. | 06:51 |
| prometheanfire | ya | 06:51 |
| prometheanfire | that's a pain | 06:51 |
| prometheanfire | I'm still probably gonna work on getting gentoo support into os-ansible, then drop the packages | 06:52 |
| ttx | yeah, I'd agree that the trade-off is not distro-friendly | 06:52 |
| prometheanfire | getting tired of all these shenanegins | 06:52 |
| ttx | basically we can keep open dep reanges on master, but not on 3 stable branches without the help of the packagers | 06:53 |
| prometheanfire | what type of help? | 06:53 |
| ttx | Stable branches were always a collaboration between upstream and packagers... and yet we struggled | 06:53 |
| ttx | fixing the stabel gate when it's broken | 06:54 |
| ttx | a new dep comes out and is not backward-compatible, breaks stable/juno. What now | 06:54 |
| ttx | someone has to debug the situation | 06:54 |
| prometheanfire | that specific dep need pinning back | 06:54 |
| ttx | sure. But someone has to care | 06:55 |
| ttx | Devs care about master and only tolerate stable branches | 06:55 |
| prometheanfire | ya :( | 06:55 |
| ttx | But then, gate maintainers are a rare resources | 06:56 |
| ttx | and if they spend all their time on stable branches, they burn out | 06:56 |
| prometheanfire | the lib update breakage could be somewhat automated | 06:56 |
| prometheanfire | do a pip freeze on every run and compare/diff between runs | 06:56 |
| prometheanfire | compare to known good | 06:57 |
| ttx | On our libraries, the switch to maintaining stable branches for all of them should be beneficial to all | 06:57 |
| prometheanfire | what's the update mechanism for updating the freeze on the stable branches? | 06:57 |
| ttx | But I agree that the "rest of the world" libraries are more difficult, and pinning is not great for distros which use system-wide stuff | 06:57 |
| prometheanfire | ya, gentoo already only just gets by with openstack | 06:58 |
| ttx | Being discussed: https://review.openstack.org/#/c/161047/ | 06:58 |
| prometheanfire | route2 for instance, some have updated it and are confused why openstack doesn't merge | 06:58 |
| ttx | Basically still discussing the bestsolution here | 06:59 |
| ttx | Feel free to jump in the discussion there | 06:59 |
| prometheanfire | I will :P | 06:59 |
| prometheanfire | we don't downgrade packages if a new version is already installed, so the os-ansible solution is better in any case | 07:00 |
| prometheanfire | and can work across distros | 07:01 |
| prometheanfire | thanks for the link, gonna comment on https://review.openstack.org/#/c/161047/ in the morning | 07:03 |
| prometheanfire | and trolling started | 07:08 |
| prometheanfire | ttx: anyway, nn thanks for the pointers | 07:08 |
| *** jamielennox is now known as jamielennox|away | 07:11 | |
| *** ihrachyshka has joined #openstack-stable | 07:22 | |
| *** ihrachyshka has quit IRC | 07:28 | |
| *** ihrachyshka has joined #openstack-stable | 07:30 | |
| *** pixelb has joined #openstack-stable | 07:52 | |
| tchaypo | prometheanfire: ttx: There was a discussion related to this in -infra this morning - triggered by a mailing list post "[openstack-dev] [all][pbr] splitting our deployment vs install dependencies” and culminating in https://etherpad.openstack.org/p/stable-omg-deps | 08:07 |
| tchaypo | it’s not precisely about what the caps should be - more about the fact that the data we currently put in requirements.txt is used for >1 purpose and we need different sets of data for different purposes | 08:08 |
| *** e0ne has joined #openstack-stable | 08:08 | |
| *** e0ne has quit IRC | 08:10 | |
| *** e0ne has joined #openstack-stable | 08:14 | |
| ttx | tchaypo: right. What we run in tests is slightly disjoint from what range actually works | 08:14 |
| *** ihrachyshka has quit IRC | 08:15 | |
| tchaypo | but there are people who want to install the known-working thing, and there are other people who need to match dependencies with other packages and would prefer to just get a probably-working range, a known-bad blacklist, and tools for testing to make sure the combination they pick works | 08:16 |
| ttx | tchaypo: ++ | 08:16 |
| ttx | There are multiple ways to protect the stable branches from direct dependency breakage that do not involve screwing up distributions completely :) | 08:17 |
| *** e0ne has quit IRC | 08:19 | |
| *** derekh has joined #openstack-stable | 08:32 | |
| *** jamielennox|away is now known as jamielennox | 09:02 | |
| *** ihrachyshka has joined #openstack-stable | 09:07 | |
| *** andymaier has joined #openstack-stable | 09:17 | |
| *** apevec has joined #openstack-stable | 09:49 | |
| *** apevec has joined #openstack-stable | 09:49 | |
| *** e0ne has joined #openstack-stable | 09:55 | |
| *** e0ne has quit IRC | 09:56 | |
| *** e0ne has joined #openstack-stable | 10:17 | |
| *** e0ne has quit IRC | 10:35 | |
| *** e0ne has joined #openstack-stable | 10:37 | |
| *** e0ne has quit IRC | 10:41 | |
| *** ttx has quit IRC | 11:48 | |
| *** ttx has joined #openstack-stable | 11:48 | |
| *** ttx has quit IRC | 11:49 | |
| *** ttx has joined #openstack-stable | 11:49 | |
| *** apevec has quit IRC | 12:02 | |
| *** jamielennox is now known as jamielennox|away | 12:12 | |
| *** ttx has quit IRC | 12:45 | |
| *** ttx has joined #openstack-stable | 12:45 | |
| *** ttx has quit IRC | 12:46 | |
| *** ttx has joined #openstack-stable | 12:46 | |
| *** eharney has joined #openstack-stable | 12:50 | |
| zul | the release announcement for 2014.2.3 isnt out yet is it? | 13:12 |
| ttx | zul: no, I think adam delayed to include trove | 13:35 |
| zul | ttx: okie dokie | 13:39 |
| *** apevec has joined #openstack-stable | 13:41 | |
| *** ihrachyshka has quit IRC | 13:42 | |
| *** ihrachyshka has joined #openstack-stable | 13:47 | |
| prometheanfire | tchaypo: thanks for that link | 14:36 |
| *** xaeth_afk is now known as xaeth | 14:44 | |
| prometheanfire | tchaypo: it still looks like no one is thinking of the packagers | 14:48 |
| ihrachyshka | prometheanfire, if there are no packagers that are actually involved in upstream stable maintainance, then it's not surprising that there is 'little thinking' | 14:50 |
| prometheanfire | do packagers know to be involved with that? | 14:50 |
| prometheanfire | for all other things that I package upstream is somewhat responsible | 14:50 |
| prometheanfire | not the case with openstack though | 14:50 |
| ihrachyshka | I don't know, I haven't seen any packagers other than from redhat or ubuntu side to be involved in any stable activities | 14:51 |
| ihrachyshka | not sure what responsible is here. there is a balance, openstack is different since it has gate, and there is shortage of resources to fix gate failures in time, hence the solution. | 14:52 |
| prometheanfire | openstack's deps move too fast | 14:52 |
| prometheanfire | or more correctly, what openstack deps on changes too much | 14:52 |
| ihrachyshka | because openstack is the largest and most active pythonic project in the world | 14:52 |
| prometheanfire | that's fine, just makes it near impossible to package fir | 14:53 |
| prometheanfire | for | 14:53 |
| ihrachyshka | for one man army, yes | 14:53 |
| *** rwsu has quit IRC | 14:54 | |
| prometheanfire | and I'm honestly fine with the dep caps, I worry that they hide actual problems though | 14:54 |
| ihrachyshka | speaking of capped deps... ideally, we would gate against min and max versions capped, plus have a bot that moves the deps forward after new pypi releases. | 14:54 |
| *** rwsu has joined #openstack-stable | 14:54 | |
| prometheanfire | ya, that bot is essential | 14:55 |
| ihrachyshka | prometheanfire, which problems? I could miss some comments | 14:55 |
| prometheanfire | if the deps are not moved forward | 14:55 |
| prometheanfire | if you attempt to move the deps forward automaticly, how do you diferentiate between a dep that was capped proactivly vs one that was capped because of a known problem? | 14:56 |
| prometheanfire | recording the known problems caps seperately could help maybe | 14:56 |
| ihrachyshka | those deps are just contract between upstream and users. what those caps say is that 'we tested the code against those versions, atm we don't have capacity to test with all versions, or newer ones, so that's all we are able to provide you; if you use other versions, you're in unsafe mode' | 14:56 |
| ihrachyshka | prometheanfire, we could mark entries with a tag | 14:57 |
| prometheanfire | those caps hide actual dep problems | 14:58 |
| ihrachyshka | there is no problem IF you use those versions, is it? | 14:58 |
| prometheanfire | not arguing that | 14:59 |
| prometheanfire | consider this scenario | 14:59 |
| prometheanfire | my distro doesn't have sphinx < 1.3 | 14:59 |
| prometheanfire | the requirements require <1.3 | 14:59 |
| ihrachyshka | prometheanfire, you can actually propose raising the cap if you feel like it | 15:00 |
| prometheanfire | how am I suposed to know if that was capped because of protectionism or if it's a cap because of an actual problem | 15:00 |
| prometheanfire | ihrachyshka: let me finish... | 15:00 |
| prometheanfire | we have the caps for two diferent activities | 15:00 |
| prometheanfire | one is protection, which is fine | 15:00 |
| prometheanfire | the other is actual known problems | 15:01 |
| prometheanfire | packagers will want to package off the second one (at least I will) | 15:01 |
| prometheanfire | if only because it has the required flexibility for distros | 15:02 |
| ihrachyshka | prometheanfire, yeah, tags or two separate lists of deps could be useful. but note that to avoid issues, it's not enough to package the thing, you still need to run integration tests, unit tests, etc. | 15:02 |
| ihrachyshka | you're packaging for gentoo, right? | 15:02 |
| prometheanfire | yep | 15:02 |
| prometheanfire | I'd like two child lists that are merged into one requrements.txt (or test-reqs) | 15:03 |
| prometheanfire | that'd solve both problems maybe | 15:03 |
| ihrachyshka | those caps can be worth for you since you're rolling | 15:03 |
| ihrachyshka | *worse | 15:03 |
| prometheanfire | I'd also standardize to overwriting the provided requirements with '' so that distros won't complain | 15:04 |
| ihrachyshka | for redhat, we just cap versions each release, and then don't move stuff forward unless there is huge need | 15:04 |
| prometheanfire | ya | 15:04 |
| prometheanfire | though I think the caps are still fine | 15:04 |
| prometheanfire | we can still roll forward and keep the old versions around | 15:04 |
| prometheanfire | [I] dev-python/requests Available versions: 2.3.0^t{tbz2} ~2.4.1^t ~2.4.3^t ~2.5.0-r1^t ~2.5.1^t ~2.5.3^t ~2.6.0^t {PYTHON_TARGETS="pypy pypy3 python2_7 python3_3 python3_4"} | 15:05 |
| prometheanfire | for instance :P | 15:05 |
| prometheanfire | so those caps are probably fine | 15:05 |
| prometheanfire | they were a surprise to me and that was painful | 15:06 |
| ihrachyshka | I see. so yeah, I agree that some distro voice is not heard, but I also think that people should be involved in stable maintainance, at least when it comes to ML discussions, because project lacks other's perspective (I haven't thought out how painful it can be for rolling releases) | 15:06 |
| prometheanfire | I still think two seperate lists are needed to track actual problems vs protection | 15:06 |
| prometheanfire | which mailing lists? | 15:06 |
| ihrachyshka | openstack-dev specifically | 15:06 |
| ihrachyshka | all decisions are done there | 15:06 |
| prometheanfire | the busy one, of course :P | 15:07 |
| ihrachyshka | caps were discussed for weeks | 15:07 |
| prometheanfire | I've tried to avoid that one because of traffic, guess I shouldn't have | 15:07 |
| ihrachyshka | and they came out mostly because we had an alternative of dropping stable branches | 15:07 |
| ihrachyshka | due to lack of human resources | 15:07 |
| prometheanfire | ya, I could have been prepared for it at that point :P | 15:07 |
| ihrachyshka | because gate was broken each week by ever new release of libX on pypi | 15:08 |
| prometheanfire | I get that | 15:08 |
| prometheanfire | and like I said, it was the surprise that was the worst part | 15:08 |
| prometheanfire | why doesn't openstack have some sort of -packagers list where we can git notified of things like this? | 15:09 |
| ihrachyshka | the usual way of doing it in openstack world is having a separate [tag] in email subject in openstack-dev@ | 15:09 |
| ihrachyshka | I actaully started one lately - [packaging] | 15:10 |
| prometheanfire | <3 | 15:10 |
| prometheanfire | openstack-dev+subscribe@lists.openstack.org ? | 15:10 |
| ihrachyshka | so you subscribe and then set filters to mark everything except interesting stuff read/move to trasj | 15:10 |
| ihrachyshka | *trash | 15:10 |
| ihrachyshka | yeah, that one I guess | 15:10 |
| ihrachyshka | http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev | 15:10 |
| ihrachyshka | prometheanfire, here is the first (and so far only) thread with the tag: http://lists.openstack.org/pipermail/openstack-dev/2015-March/059078.html | 15:11 |
| prometheanfire | meeting time | 15:12 |
| prometheanfire | lol @ web form to subscribe, but done :P | 15:22 |
| prometheanfire | ihrachyshka: and gots :D | 15:31 |
| ihrachyshka | prometheanfire, come on, don't be that 90s! email is for sissies! | 15:32 |
| prometheanfire | I'm already on about 20-30 mailing lists | 15:32 |
| prometheanfire | much bikeshedding | 15:32 |
| prometheanfire | little time | 15:32 |
| prometheanfire | http://phk.freebsd.dk/_images/bikeshed.png | 15:33 |
| prometheanfire | I kinda want that as my logo now | 15:33 |
| prometheanfire | does --config-dir go into child dirs? | 15:34 |
| ihrachyshka | afaik nope | 15:35 |
| ihrachyshka | but why | 15:35 |
| prometheanfire | just curious, I'm glad it doesn't | 15:35 |
| ihrachyshka | I can't be sure! | 15:36 |
| ihrachyshka | prometheanfire, https://github.com/openstack/oslo.config/blob/master/oslo_config/cfg.py#L1262 | 15:37 |
| prometheanfire | :D | 15:38 |
| prometheanfire | that should be one dir only | 15:38 |
| *** ttx has quit IRC | 15:57 | |
| *** ttx has joined #openstack-stable | 15:57 | |
| *** ttx has quit IRC | 16:05 | |
| *** ttx has joined #openstack-stable | 16:05 | |
| *** derekh has quit IRC | 16:57 | |
| *** ihrachyshka has quit IRC | 17:00 | |
| *** andymaier has quit IRC | 17:06 | |
| *** pixelb has quit IRC | 17:13 | |
| *** SlickNik has joined #openstack-stable | 17:22 | |
| *** pixelb has joined #openstack-stable | 17:33 | |
| *** mgagne_ is now known as mgagne | 17:58 | |
| *** tlbr has quit IRC | 18:22 | |
| *** SlickNik has quit IRC | 18:31 | |
| *** SlickNik has joined #openstack-stable | 18:31 | |
| *** apevec has quit IRC | 18:50 | |
| *** mrunge has quit IRC | 19:04 | |
| *** e0ne has joined #openstack-stable | 19:04 | |
| *** e0ne has quit IRC | 19:14 | |
| *** e0ne has joined #openstack-stable | 19:17 | |
| *** e0ne has quit IRC | 20:08 | |
| *** tlbr has joined #openstack-stable | 20:14 | |
| *** e0ne has joined #openstack-stable | 20:41 | |
| *** pixelb has quit IRC | 20:53 | |
| *** e0ne has quit IRC | 21:43 | |
| *** xaeth is now known as xaeth_afk | 21:48 | |
| tchaypo | prometheanfire: the conversation yesterday was focussed on pbr | 22:35 |
| tchaypo | and I’m told that means downstream packagers don’t need to be considered because they don’t use pbr | 22:35 |
| *** jamielennox|away is now known as jamielennox | 23:22 | |
| *** SergeyLukjanov has quit IRC | 23:39 | |
| *** SergeyLukjanov has joined #openstack-stable | 23:40 | |
| prometheanfire | tchaypo: I use pbr | 23:57 |
Generated by irclog2html.py 2.14.0 by Marius Gedminas - find it at mg.pov.lt!
