14:58:54 <richard> #startmeeting Tor Browser Weekly Meeting 2023-03-20
14:58:54 <MeetBot> Meeting started Mon Mar 20 14:58:54 2023 UTC.  The chair is richard. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:58:54 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
14:58:55 <dan_b> \o
14:58:59 <richard> it's that time of the week again
14:58:59 <Jeremy_Rand_36C3[m]> Hi!
14:59:11 <dan_b> the beginning? 😄
14:59:13 <richard> the pad as usual: https://pad.riseup.net/p/tor-tbb-keep
14:59:15 <boklm> hi
14:59:48 <PieroV> my favorite time of the week
15:01:51 <gaba> hi
15:01:58 <richard> o/
15:03:01 <richard> ok folks
15:03:26 <richard> 12.0.4 went out late last week after some unexpected though over-due android security backports
15:03:31 <richard> we'd been lucky for too long :p
15:04:03 <richard> i'm planning on having a tor-browser-build MR ready for 12.5a4 this evening and build overnight
15:04:23 <richard> and an s131 build thereafter
15:05:02 <dan_b> question: i had always thought 12.5a4, the 'a' meant alpha, as in 12.5a4 was the alpha of what would become stable 12.0.4
15:05:18 <dan_b> or is that what it means, you're cutting a tag so we can always rebuild that while HEAD moves on?
15:05:39 <richard> ok
15:05:47 <Jeremy_Rand_36C3[m]> dan_b: it's an alpha of what will become 12.5.0 AFAIK
15:05:57 <Jeremy_Rand_36C3[m]> I.e. the 4th alpha
15:05:59 <dan_b> aaah gotcha
15:06:00 <richard> so the alpha series is what will be the next major stable series
15:06:20 <richard> whereas the current stable series is basically the last of the previous alpha series + any backports from the current alpha series
15:06:21 <dan_b> any reason we jump from 12.0 to 12.5?
15:06:21 <Jeremy_Rand_36C3[m]> But I should probably defer to richard to explain better
15:06:44 <richard> right so the tor browser version goes up by 0.5 between each major release
15:07:08 <richard> so like we had 11.5 release in Q2 of 2022, 12.0 release in Q4
15:07:17 <dan_b> ah gotcha
15:07:26 <Jeremy_Rand_36C3[m]> dan_b: I think the UX/marketing people figured out that consumers like 0.5 jumps more than 0.1 jumps
15:07:29 <richard> after 11.5 was released alpha jumped to 12.0a1
15:07:37 <richard> etc
15:07:53 <Jeremy_Rand_36C3[m]> IIRC long long ago Tor Browser didn't use 0.5 jumps
15:07:55 <richard> so there's no real relatonship between the version numbers of the current stable and current alpha version
15:07:59 <richard> they just happen to have the same digit
15:08:03 <Jeremy_Rand_36C3[m]> But I don't think I was active in Tor Browser dev back then
15:08:15 <dan_b> ok that helps a bunch, thanks
15:08:20 <richard> but they aren't really meant to be in alignment or anything (just happenstance atm)
15:08:33 <richard> um ok
15:09:21 <henry-x> dan_b: good question because I didn't know why either :)
15:09:23 <PieroV> We also prefer jumping by 1 for the ESR rebase, I think
15:09:39 <PieroV> The other version is in the middle, so .5 :)
15:09:50 <PieroV> I've always thought so, at least
15:09:51 <dan_b> aaah cool
15:09:54 <PieroV> But I might be mistaken
15:09:55 <richard> ma1, PieroV: can y'all provide a quick overview of what happened last week re webRTC on windows base-browser and the current plan for that one *particular* android-backport
15:10:11 <dan_b> so 13a will start with the ESR migration work then?
15:10:14 <richard> no idea why historically we jump by 0.5, probably a question for boklm or gk at this point :p
15:10:22 <boklm> I don't know why
15:10:31 <richard> dan_b: i think that sounds right
15:10:34 <boklm> I think it's always been like that
15:10:49 <PieroV> WebRTC is quite easy
15:10:52 <ma1> I'll tell you about Android, and leave the floor to PieroV for WebRTC+ù
15:10:53 <GeKo> the major ones were for the esr update
15:10:55 * Jeremy_Rand_36C3[m] could have sworn that circa 5-10 years ago TBB didn't jump by 0.5
15:11:04 <Jeremy_Rand_36C3[m]> Maybe I'm remembering wrong
15:11:20 <GeKo> but we wanted to have kind of major ones for features we delevoped over the year ourselves
15:11:25 <GeKo> which went into .5
15:11:33 <GeKo> that's all
15:12:13 <richard> :)
15:12:32 <PieroV> ma1: do you want to go with Android first, or do you want me to go with WebRTC?
15:12:42 <ma1> PieroV, the floor is yours :)
15:13:07 <PieroV> Okay. So, we had a crash on Windows... But it was not a real memory shenanigans crash or something similar
15:13:31 <PieroV> It was a MOZ_CRASH because a promise wasn't expected to be rejected, but it was
15:13:51 <PieroV> Long story short, initialization of the IPC mechanism with the socket process failed
15:14:25 <PieroV> HTTP seems not to use it, but WebRTC does. So, the quick fix for WebRTC on Windows is completely disabling the network process
15:14:52 <PieroV> And in the meantime maybe try to send msism's great job upstream, and then discuss this problem there
15:15:31 <ma1> Re: Android. We've found one of the Android backports doesn't work on TB, and  quite unreliable on Firefox as well. Everything points at Mozilla's fix having a race condition due to the too many moving parts in Fenix/Android Components/GV.
15:16:02 <ma1> At least, that what I gathered by creating a mitigation in NoScript as a stop-gap measure.
15:16:31 <ma1> And what PieroV confirmed trying to follow the logs along many debug sheenigans (both mine and yours) on our Android tools.
15:17:02 <ma1> (which, BTW, I think we should try to put in some order as soon as we're out of other priorities like S131)
15:17:13 <PieroV> Yep, the race condition seems to be on the GV/C++ side
15:17:37 <PieroV> The Java side seems fine-ish, it just makes everything more complicated to deal with
15:17:47 <richard> ok, so mostly good news all around then
15:17:54 <ma1> Anyway, we basically shipped with this not-so-good "fix" from Mozilla, but a full working mitigation is in NoScript 11.4.19, which I've asked for expedite review in the weekend
15:18:11 <ma1> So as soon as NoScript is autoupdated TBA is protected
15:18:20 <richard> (and firefox :3)
15:18:21 <PieroV> Fine-ish and not fine because Moz's original fix ends up in an ignored error condition
15:18:29 <ma1> And in the meanwhile, we're gonna ask Mozilla to fix their fix :D
15:18:58 <dan_b> lol
15:19:28 <richard> ma1: can you keep track of that upstream and backport once/if they have something that's backportable? :p
15:19:39 <PieroV> Probably performances have been improved in more recent Firefox, so Firefox encounters the race condition less likely than us
15:19:45 * Jeremy_Rand_36C3[m] notes that it's extremely unreasonable for Mozilla to be acting as a Tivoization gatekeeper for Tor Browser security fixes that we choose to deploy via NoScript
15:19:46 <ma1> Yes, I'm cced in the bug and I'm commenting there later today
15:20:29 <ma1> Jeremy_Rand_36C3[m], something I'm trying to push for for some time now is having our own Tor-hosted NoScript update channel.
15:20:29 <richard> jeremy: android backports got harder around january of this year due to them merging fenix and android-components
15:20:38 <richard> which to be clear I think is a good thing and will make our jobs easier long term
15:20:50 <richard> but in the interim is a bit challenging to deal with
15:20:55 <PieroV> Jeremy_Rand_36C3[m]: I'm not sure I'm following
15:21:09 <Jeremy_Rand_36C3[m]> ma1: yeah that would help
15:21:23 <ma1> PieroV, I think Jeremy_Rand_36C3[m] was talking about AMO reviews
15:21:38 <Jeremy_Rand_36C3[m]> PieroV: WebExtensions need a Mozilla signature to be updated/installed, even when they're extensions that Tor Browser has already decided should be deployed
15:21:40 <richard> oh i see i see
15:21:53 <Jeremy_Rand_36C3[m]> That's an extra layer of trusted third party that shouldn't exist ideally
15:22:14 <PieroV> Oh, I see, thanks
15:22:25 <richard> agreed from me
15:22:46 <richard> dan_b/donuts: were there any issues getting the requested nightlies built/delivered for user testing last week?
15:22:53 <ma1> Jeremy_Rand_36C3[m], if we self-host a tor-specific version we can have AMO automatic signature in minutes.
15:23:03 <richard> or any issues that came out of that (not sure when the testing is meant to be happening)
15:23:13 <donuts> richard: no issues with the nightlies afaik, however there wasn't really a lot to test
15:23:15 <dan_b> shipped on our end just find, donuts seemed happy, i assume the testing went well
15:23:45 <dan_b> i'll be working through the rest of my s30 tickets until they are gone so hopefully you'll have a bit more next time
15:23:57 <richard> donuts: what's the next major deadline for user testing in April?
15:24:48 <Jeremy_Rand_36C3[m]> ma1: you mean using a Tor signing key?
15:24:49 <donuts> richard: w/c the 17th is when the final session will happen, I'm not sure on the exact day(s) yet though
15:24:50 <Jeremy_Rand_36C3[m]> (if so that seems generally sane)
15:25:14 <richard> ok good taht's not for awhile
15:26:06 <ma1> Jeremy_Rand_36C3[m], no, just uploading to AMO if you specify a non-mozilla update URL gets it signed after some automated validation for already approved extensions such as NoScript.
15:26:30 <PieroV> Well, we could even decide to have a Tor Browser key for addons
15:26:45 <Jeremy_Rand_36C3[m]> ma1: Ah. That's still an improvement I guess, but I do think requiring a Mozilla signing key instead of a Tor Project signing key is undesirable.
15:26:45 <PieroV> Nothing prevents us from doing so
15:27:21 <richard> ok
15:28:02 <richard> PieroV: can you backport any required updater patches to s131, update the updater URLs, etc for the next release?
15:28:16 <PieroV> richard: I have tor-browser!585
15:28:30 <PieroV> Once it gets approved, it should be as easy as cherry-picking it to S131
15:28:33 <richard> amazing
15:28:35 <richard> ok
15:30:05 <richard> and boklm: can you update the signing scripts for signing s131 mars for the subsequent release
15:30:17 <boklm> ok
15:30:29 <boklm> I will do that this week
15:30:45 <richard> ok
15:32:14 <boklm> by the way, did you get news about the new signing machines?
15:32:26 <richard> oh no i haven't
15:32:31 <richard> is it on the ticket?
15:32:43 <boklm> I haven't
15:33:11 <richard> oh I see
15:33:18 <richard> i'll follow up
15:34:27 <richard> ok I think that's everything I was curious about
15:34:47 <richard> i'll happily give the floor to anyone else if you have topics to discuss
15:34:50 <Jeremy_Rand_36C3[m]> boklm: any ETA on the openssl / arm-linux MR?
15:35:07 <Jeremy_Rand_36C3[m]> I think I addressed all the feedback, hopefully should be boring to review
15:35:40 <boklm> Jeremy_Rand_36C3[m]: thanks, I'm planning to look at it soon
15:36:32 <Jeremy_Rand_36C3[m]> boklm: great thanks!
15:38:07 <richard> ok, have a good week everyone
15:38:14 <richard> see you all on IRC o/
15:38:17 <PieroV> Thanks! o/
15:38:25 <ma1> o/
15:38:26 <boklm> o/
15:38:35 <richard> #endmeeting