Monero Dev Meeting Note Highlights – 2019-09-22
Logs
351 lines
rehrarok
rehrarit's time
rehrar1. Greetings
rehrarwho all is here?
sgp_hello :)
needmonero90Hi
xmrscott[m]Hey hey hey
rbrunnerHi
tevadorhey
sech1\o/
ErCiccione[m]Hi
ferretinjapanso everyone except the people rehrar pinged are here, nice.
rehrarit would seem so
rehrarmoneromooo: do we have you at least? :P
moneromoooYes.
rehrarone last round for fluffypony luigi1111 ArticMine binaryFate
rehrarplease join us ASAP
rehrarin the meantime, we can move on to 2. What's been completed since last meeting?
vtnerd__here
sech1I've finished RandomX JIT compiler for ARM CPUs. Verification time is even better than CN/R on ARM.
moneromoooPretty good stuff here.
rbrunnerNice
ArticMineHi
dEBRUYNEI am here
meksiHi
tevadorso I guess we are including the ARM JIT in the release?
tevadorprobably no reason not to
moneromoooIt can be disabled with an env var AFAIK so sure.
tevadorjust it hasn't been heavily tested as the x86 JIT
SoiMatterSirs! ferretinjapan asked moi to join.
hycsince it's easily toggled off, I'd say we run with it
jtgrassiehi
SoiMattero7 I will keep still :)
moneromooomonerod can now synced from pruned blocks (optional). I'd like this to go in before release if reviewed.
rehrarHonest question to everyone, now that CLI has reproducible builds, can we as a community decide on a date without input from Core?
selstaGUI?
rehrarthat's next selsta :P
selstayes but it would be nice to release both the CLI and GUI at the same time and that depends on core
moneromoooPony still controls the website (for binaries to be uploaded), and the DNS records (for the hashes of the binaries).
rehrarok so there's distribution and GUI to consider
dEBRUYNEPony said he'd be around for the CLI & GUI release
rehrarI mean, ultimately we want to go along this road at some point to reduce centralization anyways, no?
moneromoooAlso needs people to build for random OSes. Does Windows have repro builds too ?
ArticMineDo we have a release date in mind?
dEBRUYNEmoneromooo: Yes
dEBRUYNEAt least for the CLI
ErCiccione[m]ArticMine: i think the idea was to set a date today
rehrarso let's set one
rehrarArticMine can represent the core team here
sgp_October 26?
rehrarAre we thinking mid-October? End of October?
dEBRUYNEI would personally prefer a bit more towards the end of October, as that gives us a bit more leeway
ArticMineI do not ave an issue with setting a release date at this meeting
ErCiccione[m]i agree
hycLet's go with October 31, Halloween
tevadoris this the release date of the fork date?
tevadoror*
hycit'll be scary
moneromoooI have some reservations about the stability of randomx-in-monerod. Bugs are still coming in fast.
needmonero90Monero's spoopy fork
rehrartevador: fork date
dEBRUYNEtevador: fork date
sech1October 31 is a bit too late
moneromoooSo pushing back would be good in my view :)
ArticMineand code freeze?
rehrarif mooo has reservations then maybe 31 is a decent date?
rehrarsech1: what's your reasoning?
tevadorso better to not do it in the middle of the week
luigi1111I'm sorta here
sech1October 31 is kind of an edge of publicly promised "fork in October"
sech1October 26 sounds better
moneromoooI don't anyone (who counts) ever promised.
hycassuming it's actually ready by then
needmonero90Welcome to decentralization sech1.
rehrarI am fine with 26th
tevadoryeah, 26 October would be better if it's doable
needmonero90Community won't be too mad if we don't hit deadlines
SoiMatterGo with the date of the next UK election
hyctestnet has been disturbingly unstable the past few weeks.
hycbut most of the problems were unrelated to randomX
Inge-Better a safe launch than an october launch.
ArticMineOne key point is the time between release and fork date
rbrunnerNormal testnet or the RandomX testnet?
hycrandomX testnet
hycbut the issues with peers being disconnected and banned would have happened on public testnet too if anyone bothered to exercise that
hycbasically current git master is unstable, recent changes to the networking code broke a lot.
sech1if fork is October 26 (4 weeks), would 2 weeks be enough to fix all that?
sech1and would additional 5 days help?
ErCiccione[m]i'm starting to think would be better to posticipate to after october
tevadorso it looks more like November
ArticMineRelease and fork date ware very different
dEBRUYNEhyc: Those recent changes will be reverted by 5905 right?
moneromoooYes.
ArticMineWe need to give people time to update
ErCiccione[m]if there are so many doubts i would just wait. No reason to rush it, this is a big fork
hycagreed, safer to wait
rehrarso really all of this should have been discussed two weeks to a month ago
sgp_I don't think the optics of a "late" Oct 31 date will be terrible. We can get ahead of it with the Halloween marketing. If people prefer November, then I think we should still pick a date
M5M400will this release discontinue legacy payment IDs?
rehrarso for next time we really need to start seriously discussing the fork two months prior to "planned" date
moneromoooYes.
moneromoooThough currently not banned, just code removed.
dEBRUYNErehrar: I don't think we could have foreseen the issues arosen on testnet
moneromoooAre any large exchanges still using them ?
ArticMineI suggest an October 31 release date with a fork date 4 weeks later
luigi1111late is far better than broken
sgp_I can try to compile a list
sech1We still need to decide fork date and release schedule today
rehrardEBRUYNE: literally every single fork there is a "I don't think we could have foreseen" situation?
needmonero90What happened to move fast and break things? This is tech!
dEBRUYNEYes because there will simply be things that cannot be foreseen
dEBRUYNEYou cannot perfectly plan such a thing
luigi1111I too enjoy breaking money when possible
rehrarok, ArticMine has made a concrete suggestion
rehrar31st release date, one month later fork date
sgp_one month? why one month?
needmonero90I think its reasonable.
needmonero90No objection here.
ferretinjapanluigi1111, I agree, late and working is far better than on time and buggy…
hycis that the same offset we used in previous releases?
M5M400moneromooo: i have about 330 miners still mining to payment_id addresses. that's 5 %
sgp_no, closer to two weeks I think
ArticMineTo give people proper notice to upgrade
luigi1111that's what it ended up being
needmonero90We were pushing to the wire last time
needmonero90To be fair
luigi1111we never targeted two weeks that's simply not ideal
needmonero90Two weeks was cutting it close
ArticMineThis has been an issue many times before
ErCiccione[m]yeah, one month seems fair
sech1October 31st release and November 30th fork then?
luigi1111yes we've always pushed releases to late but doesn't mean we shouldn't try to do better
ArticMinesounds fine to me
sgp_So what specific fork date?
tevador30th Nov
rbrunnerNovember 30 is again Saturday, which is nice
sech1let's hope CN/R holds without ASICs until then :)
ArticMineSo when is code freeze?
moneromoooWhenever we don't have anything left we want to merge :)
ErCiccione[m]now with reproducible build releases should be much faster, at least for CLI. We rarely managed to release at the time we decided
needmonero90October 31st makes sense imo
sgp_I think having a release in October will help with optics. And a 31st update is easy to brand
rehraralright, it looks like we have a loose consensus, unless anyone would like to speak now
ArticMineSure but how much time is needed between code freeze and release?
rehrarin opposition
luigi1111like a week
luigi1111but fluffy fluffy fluffy
selsta+1
selstalong code freezes never work
rehrarluigi1111: did not show
dEBRUYNEWe only need fluffy to upload the bins though for the CLI release
dEBRUYNE(and do the website)
luigi1111I know but if he's building he has to be able to commit to a schedule
sgp_so Thanksgiving weekend hardfork (Nov 30)
sech1yes, better safe than sorry
hycsince we have reproducible builds we don't need him to build
dEBRUYNErehrar: He said he'd be available for the release (he'd prefer to have a timeline a bit in advance though)
rehrarluigi1111: if he's unable to commit to a schedule, then we can do the deterministic builds thing, and just make sure they get uploaded
luigi1111ok cool
luigi1111then just gooey
rehrarwhich is you
ArticMineThe time between code freeze and release need to be set buy those closest to the code
rehrarcan you commit to this schedule?
dEBRUYNEGUI gets build by pony
rehraroof
needmonero90Speaking of build privs, I would like to formally propose that some of the GUI repo privileges be transferred to dsc (the non deterministic build part at the very least)
M5M400how can fluffy still be the only person able to do some vital things?
luigi1111I'm the one going to be doing all the merges in both projects so heh
dEBRUYNErehrar: Third time and last time I will say -> He said he'd be available for the release
rehrarM5M400: that's why deterministic builds is top priority
rehrardEBRUYNE: ok
ferretinjapanM5M400, mainly because reproducible builds were never a thing before now.
needmonero90luigi1111: we really should move some of those responsibilities to dsc.
dEBRUYNEneedmonero90: I disagree, then we would have coders merging their own stuff
needmonero90He would be the new maintainer/lead of gui
needmonero90Or something
rehrarok, any other questions in regards to the set dates? October 31st release, November 30th fork.
moneromoooSo what do we need for the release ? readline fixes for repro builds, randomx. The pay-for-service stuff. Payment ID changes. Hopefully the sync-pruned-blocks. Anything else that's large or really needed ?
ferretinjapanfreeze date?
rehrarneedmonero90: that would be the first time a non-core team member will have done this. Not impossible, but certainly unprecedented and worthy of a bigger discussion which we don't have time for here in this meeting
dEBRUYNEOnce we branch the code is basically frozen right? Because then only fixes go in
ErCiccione[m]one week code freeze would be appreciated. Let's remember that there are the translation to deal with (and i beg for a code freeze since ever :D)
needmonero90Okay, tabled for now rehrar
moneromooovtnerd__: are you planning to have some more network stuff ready for the release ?
ferretinjapanI guess the #monero-pow guys would have a better idea when to freeze…
ArticMineOct 24th as per luigi1111 selsta?
luigi1111he's not the only one with access to a couple things but the only one that actively used his access
dEBRUYNEmoneromooo: The readline fix is 5892 right?
selstaArticMine: yes
luigi1111until recently where him and pigeons have both not been to active
moneromoooCould be. Depends on what people who have the problem say after testing it :)
dEBRUYNEArticMine: I think we can branch earlier, but moneromooo probably has a better view on that
ArticMinemoneromooo does Oct 24 work for code freeze?
ErCiccione[m]pigeons is active. I asked him to update the server of getmonero last time
moneromoooThat sounds plausible to me.
M5M400moneromooo: it appears bittrex and binance still use payment_id's
ErCiccione[m]and he also gave me a hand with weblate
moneromoooThanks.
vtnerd__moneromooo : yes, the split mempool is jst about done (finishing core tests nows)
rehrarwow
dEBRUYNEM5M400: Binance is working on upgrading their system, needmonero90 is in touch with them
rehrarvtnerd__: so your stuff will make it in this one?
vtnerd__and the dandelion++ remnants _shouldn't_ be too painful - although I always say that :/
M5M400dEBRUYNE: saifu. great.
needmonero90Binance indicated they should be ready by fork in October. Pushing the date back shouldnt hurt the timeline.
M5M400anyone in touch with bittrex?
dEBRUYNEThey are aware as far as I know
vtnerd__I can't say whether it will "make it in", because theres always a chance something comes up during review, etc
vtnerd__but the split mempool for the i2p/tor is close to ready. my concern is that its kind of tough the way it got shoved in
rbrunnerYou mean in a technical sense?
vtnerd__I took hyc's approach of adding a tag to the metadata, but I had to inspect a bunch of code paths to make sure stuff didn't leak
binaryFatehi all, sorry got late to the party
vtnerd__ultimately a true "physical" or "logical" split would be ideal, but it seemed kind of tough at the momnent when I tried to think about hwo to do it
moneromoooOh so it's a single txpool, just txes have a flag ?
vtnerd__yeah, unfortunately
M5M400are we aware of any GPU miner in the works for randX that will be ready in time for spork?
moneromoooThat's what I did for the sync-from-pruned :)
hycM5M400: xmrig is ready
luigi111124th and 31st targets are good with me for 30th fork
sech1M5M400 xmrig 4.0 is combined CPU + OpenCL miner
vtnerd__if thats worth a reject, then I can go back and try again, but its a little dicey otherwise (it requires a slightly larger refactor and possibly a second lmdb database)
luigi1111mostly afk now
rehrarvtnerd__: if this does make it in this time, it can be iterated on in the future no?
M5M400hyc: sech1: nice
moneromoooIt's not IMHO. I'd have done that too.
moneromoooBut then I tend to shove things in :)
vtnerd__yes, I don't think the patch makes it harder to refactor, its probably neutral from that perspective
hycI don't really see other approaches to that as being a good idea
tevadorbtw, fork height could be 1978433, which is exactly the start of a RandomX epoch and should be in the afternoon CET on Nov 30th
vtnerd__as an example of what I mean - one of the weird cases is that fluffy blocks request can leak the anonymity channel, etc, so yeah
dEBRUYNEtevador: Afternoon CET would allow most people to be online, because it would basically be late evening for Asia and morning for US
moneromoooAnd let's reset testnet to fork on an epoch boundary too ?
tevadordEBRUYNE: yeah, nice coincidence
ErCiccione[m]i guess we have a fork height. Nice
rbrunnerWith a testnet fork around release, end of October?
hycsounds good
dEBRUYNETo summarize the timeline, this is approximately correct right? Branch within 2-3 weeks, freeze October 24, release October 31, fork November 30
moneromoooI think we have everything we need for v12 now, apart from randomx, so we can fork the real testnet as soon as it gets merged.
rbrunnerEven better then
tevadorthat looks like a sensible schedule
sech1yes, plenty of time to iron out all issues
rbrunnerFamous last words :)
dEBRUYNEIt at least feels we are better prepared now than last time :-P
tevadorsomeone should strictly enforce code freeze this time
rehrarpenalty of death
ErCiccione[m]tevador: i strongly agree
dEBRUYNEmoneromooo: Do we need pony around to fork the real testnet?
PauleBertcan we reset stagenet too? it is getting quite huge
moneromoooYes, the miner's one of his nodes, and seed nodes need updating too.
moneromoooThe former's not a big problem, but the latter kinda is.
needmonero90Don't reset stagenet :(
dEBRUYNEOk, I will send him a message then
needmonero90Telegram tipbot uses it
rbrunner"Reset" means here dropping a few hundreds of blocks probably, no new genesis block
needmonero90It'll wipe all our balances ;_;
rbrunnerBack to the latest epoc date, that is
needmonero90(yes, I know, it's stagenet)
rbrunnerOr a few thousands …
M5M400are we in any way concerned about the probably 10x hashrate post fork and the amount of blocks being spit out quickly? might it be a good idea to set a ballpark initial diff?
hycI'm guessing that will be offset by the typical hashrate drop due to not everyone updating
moneromoooI am not.
sech1I don't think so. It'll rather grow slowly.
rbrunnerI think to remember from earlier forks that "doctoring" the diff is not trivial
sech1Even with higher hashrate from CPUs, hard fork drop in hashrate will make initial diff roughly the same as now
ferretinjapanI wouldn't be worried un less ASICs were currently saturating the network, but by the sounds of it, theres been no indication of that being the case…
M5M400I've seen a few big miners on the pool that are most likely CPU based. one instance of 17MH cn/r. these operations don't miss forks, as opposed to casual miners
sech1No signs of ASICs
ferretinjapanthats good.
rehraralright, is there anything else of note to discuss?
ErCiccione[m]minor: is it fine for everybody if we take down the current warning on getmonero, it's there since last fork
needmonero90We need to talk about extending the Monero Foundation's Dev tax, and the Monero name trademark issue
ErCiccione[m]i want to avoid people not reading it because they are used to it when we put up the new one
needmonero90Oh wait wrong coin
needmonero90We good
hycErCiccione[m]: good idea
rbrunnerYeah, take it down before it gets a running gag or something
rbrunner"Monero mixes up its hardforks"
rehraralright
ErCiccione[m]alright, will open a PR tomorrow
rehraralright friends, it would seem we're all good
needmonero90Is Gui considered core?
needmonero90Or is it it's own thing
hycdEBRUYNE: now that randomx has quieted down I'll get to rebasing that gitian PR of mine
ErCiccione[m]needmonero90: that's actually a good question. See https://github.com/monero-project/meta/issues/384
selstaPlease no renaming discussion.
selstaThe github issue is a better place for that.
rehraralright everyone, we'll officially break for today
rehrarthanks for coming. Good discussion, and got a lot done.
hycwho's posting meeting logs?
rehrarWe can continue on a two weeks schedule for now? Or want to go to one week mini-check in on off weeks?
ErCiccione[m]hyc: el00ruobuob, usually
hycok.
hycI think next mtg in 2 weeks is fine. after that, we can decide again
rehrarok
ErCiccione[m]unles you didn't mean on getmonero
rehrarsounds good
dEBRUYNEhyc: Thanks
dEBRUYNEErCiccione[m]: I'd prefer to update it with a general information thread regarding the upcoming fork
dEBRUYNEI can have that ready next week
ErCiccione[m]dEBRUYNE: sure! will you PR it on repo.getmonero?
IRC-Source_89guys sorry for the question , but i missed the begining of the discussion… Have you already choosen a block height for the fork?
scoobybejesustevador> btw, fork height could be 1978433, which is exactly the start of a RandomX epoch and should be in the afternoon CET on Nov 30th
dEBRUYNEErCiccione[m]: Sure
scoobybejesusthere seemed to be rough consensus on that
IRC-Source_89thank you
ErCiccione[m]dEBRUYNE: great, thank you. Ping el00ruobuob_ so he remembers to not PR it
dEBRUYNEI don't mind if he PRs it
ErCiccione[m]he usally PR them after more than one week, should be fine
dEBRUYNEAre we talking about the same thing? :P
dEBRUYNEI was referring to the new information thread regarding the fork
ErCiccione[m]I was talking about the logs :P i understood you wanted to include them in the thread.
ErCiccione[m]but anyway, we should really make a blog post about it.
sech1So will the schedule be officially announced soon?
rbrunnersgp_ was fast: https://old.reddit.com/r/Monero/comments/d7twle/tentative_monero_015_release_schedule/
hycsgp already announced it
dEBRUYNEsgp_: Perhaps worthwhile to mention in the comments why there is a bit of a delay
sgp_I don't know what happened with testnet
ErCiccione[m]el00ruobuob_ ignore what i said earlier, we need the logs :P
Sources and notes
- Meeting log: Overview and Logs for the Dev Meeting Held on 2019-09-22, getmonero.org. The official post carried the raw log and pointed readers to Hello Monero for the overview.
- Original Hello Monero URL slug recovered from the Wayback Machine index; the original page content did not survive, so this reconstruction carries the log as archived by the Monero Project.