22:59:18 #startmeeting network team meeting, 5 february 2020 22:59:18 Meeting started Wed Feb 5 22:59:18 2020 UTC. The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot. 22:59:18 Useful Commands: #action #agreed #help #info #idea #link #topic. 22:59:22 hello hello! 22:59:29 o/ 22:59:29 hihi! 22:59:36 who is here for first network-team meeting in february? :-) 22:59:46 o/ catalyst & nickm 22:59:58 our pad is at https://pad.riseup.net/p/tor-netteam-2020.1-keep 23:00:49 okay, let's start out with roadmap stuff: https://dip.torproject.org/torproject/core/tor/-/boards 23:01:02 are people doing alright there? 23:01:37 I'm pretty much wrapped up in 043 stuff the last few days, but I may get back to those things soon I hope :/ 23:01:58 yeah, that makes sense. it would be good if people could prioritize 043 stuff right now 23:02:28 i wonder if it'll just be the 3 of us here or if we should give everybody else a bit of time. usually we are more at the wednesday meetings 23:02:47 gaba is in europe so its late for her and i don't know if mike perry is home yet 23:03:14 I am here, I was trying to get through my emails :-) 23:03:43 o/ teor 23:03:48 I have been focused on Sponsor 55, I expect I will have to spend most of my time on it, at least until the end of next week 23:04:04 I still have to write the stats proposal (313) and open tickets for the essential work 23:04:48 sounds good 23:05:09 are everybody good with the stuff they need to review this week? 23:05:27 * ahf has 2 reviews that he is behind with, but doesn't seem to be blocking network team folks 23:06:01 I'll just check my reviews now 23:06:25 checking ... I'm prioritizing the ones that are in 043. I probably won't be able to review 044 code for a little while 23:06:41 oki, #32373 is the only one that has no reviewer and is from a NT person 23:06:46 err 23:06:55 #32372 23:07:18 ok, everything is assigned now. nice 23:08:04 I wonder if we have too many reviewers on #33072 and #33029? I'm not exactly sure who is supposed to be doing what when. 23:08:10 hey 23:08:16 ohhhh, I thought it was in an hour 23:08:18 nice 23:08:28 gaba: :-D then i'd be very tired 23:08:33 mmm, nevermind, confusing meetings... 23:08:34 hehe 23:08:40 yeah, s55 is in 50 min i think 23:08:41 s55 is in one hour... 23:08:43 yes 23:08:55 one way or the other, I'm here :) 23:08:56 nickm: hm, that is a good question. i haven't followed it much myself 23:09:02 but is it one of those things with too many chefs in the kitchen or? 23:09:05 I just hope nothing is blocking on me 23:10:04 I guess should have a look soon :/ 23:10:39 ah, it's david who wrote one of the patches 23:10:49 i was about to say that this seems like the kind of stuff david is usually very interested in 23:11:25 nickm: anything we need to delegate to somebody else? 23:11:57 I have commented on those tickets, and removed myself as a reviewer. 23:12:01 yeah, saw that 23:12:20 Can I delegate #33069 to asn? it's related to something he's working no 23:12:24 *working on 23:12:46 i can try to talk to asn tomorrow about it 23:12:52 ok 23:12:54 asn can also see the backlog 23:12:59 but i'll note it down 23:13:02 cool! 23:13:25 let's do a 0.4.3 status: https://trac.torproject.org/projects/tor/wiki/org/teams/NetworkTeam/CoreTorReleases/043Status 23:14:15 * ahf is going to focus on #33119 tomorrow 23:14:23 I've got a lot of stuff there. Several of them are more advanced than it looks like there, but there's more to do 23:14:49 hm, arma has a ticket there. i hope he is aware of that 23:15:23 I can take #33103 ; I looked at it for a moment and the fix looks simple 23:15:35 is this the one catalyst found last week? 23:15:48 hm, no 23:16:32 I just did a batch modify to put the "must" tickets (and only the musts) as "high" priority 23:16:33 no, that was getconf 23:16:38 yeah 23:18:15 anything else for 0.4.3? other than people should prioritize this since we have a lot there for the next few weeks with an aim of medio april 23:18:50 specifically we're trying to get the stable out at April 23:18:57 so we really need an rc in March 23:19:20 yeah 23:20:16 ok, we have some discussion items: 23:20:24 * What are our next steps with the C style work? 23:20:32 i think it's added by nickm 23:20:38 (i'm asking that becasue catalyst and teor are both here) 23:21:15 What do you need to know, nickm ? 23:21:27 so, it seems that we've got three areas of work to do: 23:21:39 1) pick out what style target we want and get all the settings right 23:22:00 2) do the necessary cleanup and infrastructure so that we can run that style without breaking cocci and/or checkSpaces 23:22:14 3) get the infrastructure together to automatically apply it for everybody 23:22:44 Right now catalyst is reviewing my changes on 2; we've been discussing 1 a little, but we don't have a firm consensus 23:22:46 i'm still very nervous about applying the new style to the entire tree at once. it'll be harder to review 23:23:20 (and 3 is stalled till we get farther on 1 and 2) 23:23:25 i sent out a poll about some formatting options where we don't seem to have consensus, but only see responses from teor? 23:23:43 where / when did that go? 23:23:59 (also i still think clang-format is too aggressive about reformatting) 23:24:21 nickm: Jan. 21 to the network-team list 23:24:26 does any of the other tools do more like what you'd like? 23:25:26 i'd have to go back and check... clang-format did seem the least bad, apart from being overly aggressive about rearranging line breaks and comments 23:25:33 i think you 3 should feel free to not worry too much if people don't respond to your polls. people have had plenty of time to suggest changes here 23:25:43 I don't see a poll? 23:26:18 nickm: Re: [network-team] Yet more fun with clang-format 23:26:52 The one starting with "Bumping this again." ? 23:26:54 it was more like explicitly proposing two specific changes to to nickm's configuration 23:26:57 nickm: yes 23:27:21 Ah. I didn't see that as a poll 23:27:35 I had a strong opinion on one of them, so I replied. I didn't mind what we chose for the other topics. 23:27:57 I don't feel strongly about either of those options 23:28:01 also unresolved is if we choose clang-format, what minimum version do we use? 23:28:27 I think people may have more opinions when they actually see the consequences of reformatting 23:28:45 So let's get something working, at least as a demonstration? 23:28:49 how do we make progress here? 23:28:50 Then we can iterate 23:29:04 teor: there's already the demo branch that nickm made. are people not looking at it? 23:29:08 don't we have something that we can see? i have seen a version of the tree with this applied in a tarball? 23:29:36 nickm's demo branch doesn't work yet: it doesn't take my new code, and automatically reformat it. 23:29:54 That's been our experience with CI: people have opinions once they actually see it working on their code. 23:30:20 #32931 is the last work I did here 23:30:24 err no 23:30:27 i guess another thing people can do is try editing a few files on the demo branch and see how hard their editor fights with the reformatted code 23:30:28 #32921 23:31:03 note that there are two branches there: one runs the reformatting on the code as a demo, but the other one is just code cleanups 23:31:18 so there is a missing step where clang-format is applied on ... travis? and then we see if there is a delta between what is submitted, or? 23:31:20 demo_clang_foramt_20190110 is the demon branch 23:31:45 demo 23:31:55 [clang_format_prep_2 is the one with only the code cleanups] 23:32:57 so when i last had to bulk-reformat a source tree, we wound up doing a thing with some custom elisp, because everything else we looked at was worse. it would be nice to be able to avoid that 23:32:59 clang-format should be applied as a pre-commit hook, and modify the code that's about to be committed 23:34:03 We may also want to apply it as a pre-merge hook, which adds a commit that reformats the code, before merging to master 23:34:17 Then we can merge forward succcessfully 23:34:18 i think pre-commit hook or folks editors' save-file-hook, but i think the CI should also check if the style is what the CI expects 23:34:23 Are there suggestions for what I should do next on this? 23:34:39 Let's choose a small part of the tree, based on: 23:34:58 we moved to clang-format with irssi 2 years ago and we just had a date where we said we move on that date and then we broke all the PR's from before that. i was happy with that mostly because it distributed the load on fixing things to everybody and we knew upfront when the style change happened 23:35:15 a declared reformat date seems smart 23:35:24 * code that check-spaces and coccinelle are happy with after reformatting 23:35:41 * people who are working on that code who are happy to deal with reformatting issues 23:35:54 Then we can try reformatting that small part of the tree automatically? 23:36:14 And if that works, we can declare a date for the whole tree> 23:36:15 ? 23:37:34 another suggestion: * the postprocessing script should be as small/simple as possible, to minimize our ongoing maintenance burden 23:38:12 hm. is it too complex as it stands? 23:39:05 haven't looked recently. it might be too complex if we add code to deal with making aligned escaped newlines in macros emacs-consistent? not sure 23:40:03 I would be happy to deal with reformatting in Sponsor 55, so it might make sense to target the relay, dircache, or dirauth modules 23:40:24 i like being able to do C-c C-\ (or whatever it is) to align escaped newlines in emacs, but maybe i can figure out an emacs config that produces what clang-format does with AlignEscapedNewlines:Left 23:40:51 david and i spend some time a year or so ago and took all our common tor vim knowledge and put into a vim module/package 23:41:00 i think that was useful. maybe the emacs folks should do something like that 23:41:15 we found out we both had a lot of smart editor tricks we used, but not much overlap in them. now we share them 23:41:20 ahf: oh cool. how complicated was it compared to what is needed for other C source trees? 23:41:48 we did a dialect of C in vim called "onion C" which the script detects when it is in tor.git and enables that dialect 23:41:57 ahf: nice 23:42:07 teor: I don't think we're going to have "check-spaces and cocci happy after reformatting" there necessarily. There were little changes of various kinds that needed to be made for #32921, and check-spaces stuff I had to remove because it conflicts with our new style 23:42:08 like the two spaces and our one space label: things for goto's and such 23:42:24 https://github.com/ahf/onion-vim 23:42:24 i might work on an emacs C style for tor; maybe that would help 23:43:07 are there emacs users who would like to collaborate on C style configs for emacs? 23:43:20 are we targetting the current style or the new one? 23:43:56 i'd say the new one is more important, but both would be nice 23:44:00 i can show you want I have now for emacs, but i don't know how much that would help with the new format 23:44:02 I use emacs, I'd be happy to try a style 23:45:06 i think doing knowledge sharing for this kind of stuff was very useful for david and i 23:45:14 teor nickm: cool, thanks 23:45:39 teor: i think we still disagree about SpaceAfterLogicalNot? anyone else have anything to say about it? 23:45:56 My opinion is htat once we're running clang-format, it would be easy to turn that option on or off 23:46:07 * ahf is cool with either :-) 23:46:35 i will note that making it true turns "if (!!foo)" into "if (! ! foo)", which to me seems particularly ugly and breaks up the "cast-to-boolean operator" 23:47:25 I don't mind it that much, but I will note that the single ! is much more readable with a space after it. 23:47:46 Perhaps we can do a "! ! " to "!! " post-process? 23:47:47 we can turn "! !" into !! easily 23:48:02 hm, i think i only see that being used 3 places? 23:48:05 the !! 23:48:15 ah, no 23:48:28 never mind 23:48:54 how can we progress forward on this? 23:48:57 teor: hm, how would you compare readability of, say, (!foo) vs (!(a && b)) without the space? how about (!FOO)? 23:49:43 I honestly really don't care that much 23:49:49 (10 min. left before another meeting begins) 23:50:22 I would rather prioritise getting auto-formatting working, than spend a lot of time deciding on the exact details 23:50:52 I also agree with nickm: once we are auto-formatting, changes are easy 23:50:57 i also don't think we will ever reach something where we all agree 100% about the settings 23:51:21 i have many opinions about the style, but i also know that i very quickly will adapt to whatever we pick and based on what i have seen so far it is very small things that would "bug me" in the beginning 23:52:04 i still place a lot of value in "not too different from a widely-adopted style" 23:52:30 we're closer to knf than we were before, and we can get closer 23:53:10 I'm much more interesting in getting auto-formatting working, because I think we will get style feedback from more people once it actually affects them 23:53:58 catalyst: yeah 23:54:04 Does this block review of #32921 ? 23:54:19 nickm: does which block review? 23:54:52 getting the exact style we want and getting auto-formatting together 23:55:20 teor: the items you have in the discussion is largely read/only for some specific folks and then getting feedback from your proposals, right? 23:55:23 ok online 23:55:30 * asn reads backlog 23:56:11 (should we have a regular time to talk about style and formatting?) 23:56:39 i think that would be good 23:56:46 (I can also break 32921 into smaller branches if that's helpful) 23:56:52 nickm: it seems like patch party time might be good for that, given that teor has some opinions? 23:57:17 so, this time, but 24 hours earlier than this meeting? 23:57:18 ahf: yes, I think people can do their reviews, or hand them over, once they get reminders 23:57:18 but do note that everybody in the team thinks the work is good and will say otherwise - the silence from other people should not be seen as a blocker i think (based on talking with people in 1:1's) but rather than people don't have very strong opinions about this 23:57:31 nickm: No, this exact time 23:57:59 oh! I could do that for a few weeks, yes, but it does make my dinners pretty awkward. 23:58:00 https://trac.torproject.org/projects/tor/wiki/org/teams/NetworkTeam#MeetingsSchedule 23:58:13 teor: yes, i meant at the actual patch party slot 23:58:14 Wednesday 2300 UTC 23:58:29 ok. let's do it for a few weeks and see how much progress we can make? 23:58:35 this time and monday is the best times for everybody i think. most people show up for those two timeslots 23:58:36 Great! 23:58:47 We need to start Sponsor 55 soon 23:58:57 * the Sponsor 55 meeting 23:59:10 yeah, please see the last items on the discussion list by teor and help them review their proposals 8) 23:59:19 thanks all for joining! 23:59:25 #endmeeting