19:01:05 #startmeeting 19:01:05 Meeting started Mon Nov 17 19:01:05 2014 UTC. The chair is GeKo. Information about MeetBot at http://wiki.debian.org/MeetBot. 19:01:05 Useful Commands: #action #agreed #help #info #idea #link #topic. 19:01:21 who is here? 19:01:34 <- lurking 19:01:51 * boklm is here 19:02:00 I am semi-here 19:02:02 * arthuredelstein as well 19:02:59 okay, then it is probably best to start with semi-mikeperry I guess :) 19:03:15 I accidentally double-booked another meeting now due to timezone confusion 19:03:40 Last week, I wrote patches for #13742 and #13746, and continued the endless 19:03:40 rebuild struggle with 4.5-alpha-1. The latest build (tag tbb-4.5a1-build2) 19:03:40 should be uploaded to 19:03:45 https://people.torproject.org/~mikeperry/builds/4.5-alpha-1/. 19:04:09 I also wrote the blog posts for the release: https://blog.torproject.org/blog/end-life-plan-tor-browser-32-bit-macs and https://blog.torproject.org/blog/tor-browser-45-alpha-1-released 19:04:59 This week, I plan to finish up the 4.5-alpha-1 release and write our year-end report. I also plan to review http://www.chromium.org/Home/chromium-security/client-identification-mechanisms and send the authors a mail. 19:06:03 I think that's it for me. I am also going to be distracted for most of the rest of this meeting, due to the double-booking issue 19:08:50 here is what I did: 19:09:46 I tested/invesitgated/fixed issues wrt the 4.5-alpha-1 release, namely 19:10:00 #13742, #13751, #13776 19:10:44 then If xied the LXC + Linux builds + libgmp related work which is ready to get merged 19:12:49 I wrote a patch for the remaining build related Firefox upstream issue 19:13:05 but that breaks on try, so I have to find a better solution 19:13:30 then I started integrating the first feedback for the security slider 19:13:56 and I gave #13439 a short look 19:14:30 next week I plan to: 19:14:39 finish the Mozilla patch 19:14:57 get more feedback wrt to the seucrity slider integrated 19:15:14 test the fix the EFF has for us wrt to the sqlite issue in #11630 19:16:03 get a proper Tor Browser signing key created and distributed 19:16:47 yeah, that's it so far 19:17:04 What are you going to do on Tuesday - Friday? 19:17:05 ;) 19:17:18 not sure yet, we'll see :) 19:21:14 who's next? 19:21:36 * arthuredelstein can go next 19:21:46 Last week I wrote a patch for #13672 19:21:59 And I nearly finished a patch for #13671 19:23:03 This week I'll finish that up and work on fixing unit tests for our Firefox patches and looking at what can be upstreamed 19:23:17 That's it for me 19:23:34 you might have some insight into #13749 19:24:17 I think the only thing I really don't want is Mozilla complaining about the wrong unittest framework 19:24:19 Yes, I'd be happy to work on that if no one else already is 19:24:40 maybe mochitest is better maybe xpcshell maybe there are things we should split, dunno. 19:25:14 I think xpcshell tests might be fine for the 19:25:21 *these 19:25:57 I thought so, too. So, yes, if you could work on that bug that would be great 19:26:03 OK, will do 19:26:23 * MarkSmith Can go next (once discussion dies out). 19:27:03 Last week Kathy and I almost finished the core browser patches for #13379. 19:27:11 We fixed the last bug earlier today ;) 19:27:18 We are doing some final testing and will post the patches for review later today (some backported Mozilla patches plus some of our own). 19:27:27 This week we will work on integration with our build system. 19:27:33 yay 19:27:34 Signing MARs requires running one signmar command for each MAR file that is to be signed… which makes for a lot of password prompts. 19:27:42 Mozilla's build team uses a Python script that asks for the NSS certificate database password once and pipes it to a bunch of sign mar commands. We will want to do something similar. 19:27:49 This week we will also help sort out #13776 (incremental MARs not reproducible). 19:28:01 That's all for us. 19:29:58 * boklm can go next 19:30:21 Last week I've been mainly working on tor-messenger 19:30:51 On the auto-rebasing on gecko-dev part, I did the changes Mike suggested last week (ignoring more context lines), which allows to apply a few more patches 19:31:09 they are listed as warning on this page: https://people.torproject.org/~boklm/tmp/tests/r/S2kNQApISx/browser-rebase.html 19:31:24 that's all for me 19:32:04 hmm... that's still a lot of broken patches :( 19:32:23 yes 19:33:26 I think they will require manually fixing the conflicts 19:35:31 boklm: do you plan something for this week or is this another one working mostly on tor-messenger? 19:36:10 (lagged out lemmie know if I missed anything important) 19:36:38 I think I will continue looking on the next things to do about auto-rebasing, but will mostly be on tor-messenger 19:36:46 ok 19:37:18 * tjr can give a update at some point 19:39:29 boklm: what do you think would be worthwhile doing in this regard? 19:39:47 should somebody start rebase the failing patches manually? 19:40:15 I still fear this is kind of a huge time sink if we try to keep up with Mozilla's work 19:40:48 s/rebase/rebasing/ 19:43:47 It is difficult to know what is the best approach… maybe "fix the patch now / as we go if the fix is trivial but wait if it looks complicated" ? 19:44:15 yes, maybe we can start with only the easy ones 19:44:32 sounds at least like an approach to try, yes 19:45:58 ok, tjr? 19:46:37 I have PartitionAlloc building in 4.5 alpha for Mac and Linux (without any partitioning yet). 19:46:40 I also talked with this guy: http://guilherme-pg.com/2014/10/15/Partitioned-heap-in-Firefox-pt1.html 19:46:52 Turns out jemalloc3 with a Partitioned Heap is likely to come to FF in time for ESR-next. (No promises though.) This would mostly invalidate any PartitionAlloc work, although some of it (e.g. a random+stack-framed based partitioning scheme) would likely not go into FF but may be useful for TBB. 19:46:57 Anyway, my builds of PartitionAlloc busts because memalign is not implemented in PartitionAlloc. 19:47:01 (I tried to figure out how detached debug symbols work, but failed. No matter how I interacted with them, I got better debugger output _without_ using them.) 19:47:04 I’m going to have to try and implement memalign myself, but it makes me pretty nervous because there’s a high risk of me busting the internals of the memory allocator in subtle and difficult ways to detect. 19:47:08 So there will be some considerable testing happening to feel good about it 19:47:10 I’m not sure if anyone has fancy setup for slimming down the build process for testing & development purposes, but I may try and document some techniques for that and/or submit patches. 19:47:14 Finally, not TBB, but I got my ‘run your own tor network’ blog post up: https://ritter.vg/blog-run_your_own_tor_network.html 19:49:10 how dows your current setup look like? 19:49:57 I have an ubuntu box I build stuff with, slowly and painfully, not using LXC. I'm thinking about trying to hack things to do something like... 19:50:12 a) build FF outside the vm for non-determinstic but faster testing 19:50:26 b) let you choose if you want to build 32 _and_ 64 bit linux or just one 19:50:43 c) save some files places for faster mac/win builds 19:50:47 d) not sure what else 19:51:23 what I do is just building outside the VM and then copying the build over a Tor Browser bundle 19:51:29 What do you mean by c) ? 19:51:33 that works pretty well and is quite fast 19:51:49 brade and I have also done the "build outside and copy over" trick 19:52:00 when everything is ready the final test is the one inside gitian 19:52:31 But something integrated with the build system would benefit all of us and make that kind of thing more reliabk,e 19:52:36 umm… reliable. 19:53:04 Okay, that's what I figured - because I'm only doing this in a couple hours in the evenings, some nights a week, I can do 12 hours builds 19:53:22 If I was working more full time I'd have found something better. I may try that and documenting it 19:53:32 And/or integrating it into code 19:56:08 tjr: might be worth updating #10281? 19:56:19 and nice work btw! 19:57:00 Yea I will do that, and get my patches up onto my repo on github 19:57:18 (Or if preffered, I can request repos on tpo) 19:58:54 do we have anyone from the support team here? 20:01:03 does not seem to be the case. anything else for the meeting? 20:03:51 thanks everyone. and without Mike's fancy equipment this is just an #endmeeting 20:03:52 not much to say >.> 20:03:59 #endmeeting even