13:29:44 <nickm> #startmeeting
13:29:44 <MeetBot> Meeting started Wed Apr 29 13:29:44 2015 UTC.  The chair is nickm. Information about MeetBot at http://wiki.debian.org/MeetBot.
13:29:44 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
13:29:47 <nickm> Greetings!
13:29:50 <nickm> Let's start with updates.
13:29:52 <nickm> I'll go ahead:
13:30:01 <nickm> I've been merging code and fixing minor stuff and answering questions...
13:30:18 <nickm> but not writing much code, because ED transition planning has got me pretty swamped.
13:30:21 <nickm> I hope that changes soon!
13:30:45 <nickm> I'm excited about lots of the stuff I'm meaning to do in 0.2.7, and about lots of the stuff I've been merging.
13:31:17 <nickm> Hackingwise, Coming up next I want to knock over a couple of big 0.2.7 tickets, and help isabela with ticket management however she asks me to.
13:31:27 <nickm> I'm open to sggestions for what I hack on.
13:31:30 <nickm> Who's next?
13:31:35 <teor> I've been out of touch, is there a date/feature deadline for 0.2.7?
13:31:42 <Yawning> nickm: I can go
13:31:53 <Yawning> #6411 was merged (yay)
13:32:16 <Yawning> I worked on root causing #15816
13:32:37 <athena> hi meeting
13:32:51 <Yawning> next up on my slate is to catch up on e-mail/paperwork and do pt stuff
13:32:55 <Yawning> and tor stuff related to pts
13:33:02 <Yawning> (and everything else as needed)
13:33:14 <Yawning> think that's it
13:33:15 <nickm> grand
13:34:03 <nickm> who's next?
13:34:18 <teor> I can go
13:34:48 <nickm> go for it
13:35:18 <teor> I got #15642 merged, which fixes a bug where testing networks could have contacted public fallback directories
13:35:33 <teor> (But we don't have any fallback directories yet_
13:36:04 <teor> #15771 was identified as an issue on the mailing list, and so I diagnosed and fixed it
13:36:40 <nickm> kudos on those
13:36:45 <teor> (For people who want to run a private network without TestingTorNetwork, for some reason)
13:36:53 <teor> Minor fixes, but important
13:37:18 <teor> Last few days I've been shoving gigabytes of data through test networks using chutney
13:37:28 <nickm> learn anything interesting?
13:37:51 <teor> Hidden services don't handle concurrent clients well
13:37:58 <nickm> any idea why not?
13:38:26 <teor> They appear very janky, long hangs for several seconds, unlike every other type of connection at that load
13:38:45 <teor> I haven't checked the log, but it suggests resource/locking contention
13:38:50 <nickm> several seconds?!  That's pretty weird.
13:38:55 <Yawning> shouldn't be locking
13:39:01 <teor> on localhost, too
13:39:21 <dgoulet> FYI, we have a couple of tickets detailing bottle necks on high load HS
13:39:23 <teor> I'll need to check logs, Yawning, just a first guess
13:39:32 <teor> Yes, I commented on one of them
13:39:41 <teor> It was what gave me the idea
13:39:42 <Yawning> yeah, the bulk of the code is single threaded
13:39:48 <Yawning> which is why I'm going to say it's unlikely :P
13:40:18 <nickm> logs could be one way; profiling could be another
13:40:34 <teor> Yeah, all very possible when everything is running locally
13:40:35 <dgoulet> we have nice profiling trace of uber busy HS
13:40:50 <dgoulet> teor: could be neat if you can match those with your experiment :)
13:40:53 <Yawning> oh that's right, working on that crypto stuff
13:40:54 <teor> I can get a nice trace and then see what happens when I change things
13:41:01 <Yawning> also was on my todo list
13:41:14 <Yawning> teor: do you open a bunch of connections then shove a ton of traffic?
13:41:21 <teor> Yes.
13:41:31 <Yawning> so it shouldn't be affected by the curve25519 stuff
13:41:48 <teor> There is a recent chutney feature which allows testing HS connectivity as well as client and bridge clients
13:42:06 <teor> I don't know about the curve stuff - is there an issue?
13:42:12 <Yawning> kinda
13:42:17 <Yawning> there's like 4 or 5 tickets
13:42:23 <teor> I just upped the data sent from kilobytes to gigabytes
13:42:24 <Yawning> that boil down to "curve25519 is slow"
13:42:31 <Yawning> that I was going to look into
13:42:36 <nickm> well, slower than it needsto be
13:42:41 <nickm> faster than RSA :)
13:42:42 <Yawning> right
13:42:56 <Yawning> faster than standard DH too
13:43:03 <teor> And inverted the connection logic so each client connects to one HS, rather than one client connecting to each HS
13:43:18 <teor> Anyway, those changes are slowly being packaged up into trac tickets
13:44:08 <teor> #14174 chutney verify: make amount of data sent configurable (Tor load / performance testing)
13:44:30 <teor> #14175 chutney verify: report client, bridge, and HS performance
13:45:13 <teor> I am happy with the code, but nickm and I discussed chutney at the patch meeting
13:45:23 <teor> Not modular enough, so adding features is messy
13:46:08 <teor> So I'll just do basic performance testing now, and leave additional work for chutney 2.0
13:46:18 <teor> Anyway, two more quickly:
13:46:39 <teor> Review #14882 as I wrote the original code for two of the option
13:46:56 <teor> And wrangle the clang sanitizers, again
13:47:03 <teor> #15817
13:47:09 <teor> done
13:47:26 <nickm> who's next?
13:47:33 <Yawning> Oh, I forgot to mention
13:47:51 <Yawning> 4th throuth 6th are bank holidais here
13:48:06 <Yawning> I plan to sit on my ass, drinking beers
13:48:24 <Yawning> unless everything catches on fire
13:49:18 <nickm> neat, what kind of holiday?
13:49:45 <Yawning> random holidays for the sake of having a few days off mid-year
13:49:54 <nickm> cool
13:50:02 <nickm> athena, dgoulet  -- one of you want to go next?
13:50:06 <dgoulet> sure
13:50:26 <dgoulet> bunch of tickets, currently making a revision on #15816
13:50:44 <dgoulet> did work on #15801, that made two tickets
13:52:47 <dgoulet> that's basically it on little-t tor side, I have some other questions for which I need to open tickets maybe
13:53:07 <nickm> ok
13:53:16 <nickm> athena: you ready?
13:55:51 <nickm> isabela: you want to go?
13:55:58 <nickm> anybody else?
13:56:06 <nickm> or we can move on to discussion till they're back.
13:56:47 <isabela> oi
13:56:48 <isabela> yes
13:56:49 <isabela> sorr
13:56:52 <isabela> sorry
13:56:54 <nickm> np
13:56:56 <isabela> 2 things! :)
13:57:53 <isabela> 1. we have two months (may and june) till july when we said we would freeze 0.2.7 / once I am out of the reports pile I will look deeper into the 0.2.7 situation and start tracking it more closer on these meetings
13:58:40 <nickm> it is not end-of-world if we decide to adjust the freeze date, but we should do so with careful thought.
13:58:43 <isabela> 2. roadmaps! I will come back to them as well and bug all the teams not only core tor :) but that said, random items for april which I think is ok to drop. we should look into it and reorganize it.. september is huuuuge
13:58:49 <isabela> things can be moved around
13:59:00 <isabela> I will be pinging more next week on that as well
13:59:02 <isabela> <done>
13:59:14 <nickm> athena: you back?
14:00:06 <nickm> athena: please fill us in when you're back.  For now, discussion?
14:01:47 <teor> isabela: it would be good to know how many tickets we have to implement / reivew / merge per day until freeze
14:01:49 <athena> *sigh* damned useless week stressing out apartment-hunting and moving, tbh.  finishing my review of nickm's 12498 branch is next on the agenda once that settles.
14:01:58 <teor> when you have time
14:02:08 <teor> #12498
14:02:14 <dgoulet> nickm: what's the plan for DirAuth to deploy #15850 ?
14:02:37 <nickm> athena: would collaborating on that help?
14:02:42 <nickm> or posting whatever review you already have?
14:02:58 <athena> yeah, i'll post a partial one soon
14:04:15 <nickm> great
14:04:34 <nickm> I've got to hunt down one unit test issue in that branch, but I'd like to make progress on it since other stuff depends on it.
14:04:52 <nickm> Also, a couple of weeks back you said you'd done these two tickets. I merged one but didn't see an actual branch for the other
14:05:41 <nickm> need me to find the number?
14:05:52 <nickm> meanwhile, issues for discussion?
14:05:57 <nickm> the big one for me is "what do I work on next"?
14:06:27 <dgoulet> 10:03 < dgoulet> nickm: what's the plan for DirAuth to deploy #15850 ?
14:06:28 <dgoulet> :)
14:06:56 <nickm> dgoulet: usually they update one at a time, and/or when new releases are out
14:07:13 <nickm> it's not crazy to think about getting an alpha and maybe another 0.2.6 out soon
14:07:18 <nickm> we've sure merged a lot of stuff
14:07:56 <dgoulet> would be neat, as long as DirAuth can stop voting for bad HSDir would be great
14:09:22 <nickm> agreed
14:10:03 <isabela> do we need to kill 0.2.3?
14:10:25 <nickm> we should probably kill everything older.
14:10:38 <nickm> I think that we should have aplan for killing 023, since we've said it doesn't get any more fixes
14:10:42 <nickm> (we've said that, right?)
14:10:57 <isabela> is on the roadmap to kill it that is why I am asking
14:11:14 <teor> die die die
14:11:19 <nickm> we should make a plan and officially deprecate it.  If we don't execute the plan till later, that's fine
14:11:47 <isabela> #15233
14:12:04 * nickm nods
14:12:09 <isabela> ok
14:12:49 <teor> Do we know how much of the network is still 0.2.3?
14:12:52 <nickm> I'm thinking of taking another look over mergeable stuff, then working to get an 0271 alpha out, then trying to get mvdan's thing merged
14:13:09 * mvdan pops head out
14:13:18 <mvdan> I have a few things to do in that regard
14:13:26 <nickm> cool; timeframe on those?
14:13:30 <nickm> teor: https://metrics.torproject.org/versions.html
14:13:42 <mvdan> nickm: one, two weeks maybe
14:13:47 <nickm> ok, then I can hold off a little
14:14:01 <mvdan> I also ran clang's scan-build on my code, nothing found which was weird
14:14:02 <nickm> do you have any questions on that, or do you think you're likely to think of some?
14:14:05 <nickm> great
14:14:22 <mvdan> I had a couple of questions pending in the trac ticket iirc
14:14:25 <mvdan> in the last comment
14:14:39 <nickm> then we'd better look at them; hang on...
14:14:58 <nickm> (ug, up to 141 tickets in the 0.2.7 milestone? I got us down to 130 last week!)
14:15:12 <mvdan> I've been a bit absent these past weeks, sorry about that
14:15:24 <isabela> !
14:15:25 <Yawning> HSPOST would be nice for alpha
14:15:38 <Yawning> if DonnchaC doesn't fix the branch, I can
14:15:46 <Yawning> and get it mergable
14:16:00 <teor> mvdan, I run clang analysis on the tor codebase regularly
14:16:15 <mvdan> teor: I was wondering why there were only a few dead assignments :)
14:16:32 <mvdan> also, running scan-build was not trivial, the makefile that auto* generates has a hard-coded -std=gnu99 in CC
14:16:44 <mvdan> so you have to add that std flag to CFLAGS for scan-build to work
14:16:55 <teor> I am slack and use Xcode
14:17:12 <nickm> mvdan: just answered
14:17:15 <mvdan> nickm: thank you!
14:17:19 <teor> But patches to do that with the makefiles would be great, mvdan
14:17:43 <mvdan> teor: so far I've gone the dirty way, will try and do a patch
14:21:04 <nickm> anything else for the meeting?
14:22:10 <teor> nickm: have we killed 0.2.3 already?
14:22:22 <teor> The last metrics report containing 0.2.3 is 27 March
14:22:33 <nickm> teor: We may have disallowed servers with it; let me check...
14:22:59 <nickm> we disallow servers before 0.2.4.18-rc
14:23:01 <nickm> so that's good
14:23:06 <teor> yes, from 27 March
14:23:07 <nickm> clients are still around though, maybe
14:23:09 <nickm> probably
14:23:47 <teor> yes, we still permi "0.2.3.24-rc, 0.2.3.25"
14:25:30 <teor> after I clear 5 tasks this week, I need to work out what to do next
14:25:43 <teor> Are isabela's queries going to help with that? :-)
14:25:48 <nickm> possibly!
14:26:03 <nickm> if you want fun stuff to do and you look in the 027 milestone, that might rock
14:26:26 <teor> Yeah, that's how I picked out #14882 for review
14:26:48 <dgoulet> nickm: oh before we wrap up the meeting, let me know if you need my help on some little-t tor stuff, I've been focusing heavily on R related stuff lately (thus lots of HS)
14:28:04 <nickm> We'll need some help, yeah.  Maybe to offer to help athena with that code review from hell, if she can use help?
14:28:19 <nickm> and/or look at stuff in 027 with high priority and see what excites you
14:28:29 <dgoulet> sure
14:28:31 <teor> nickm: it's not urgent, but in a week or two, can we do a chutney review / merge party? Got 4 branches I'm working on.
14:28:56 <nickm> sounds fun
14:28:59 <nickm> #endmeeting