Monero Dev Meeting Note Highlights – 2019-09-08
Logs
351 lines
rbrunnerLooks like meeting will start in a few minutes. Stay tuned.
hyctime?
hycwho's already here?
moneromoooWell, it can start now. Who wants to say anything ?
vtnerd__present
jtgrassiehola
rbrunnerHi
moneromoooI've been working on a "sync pruned blocks" patch, it's proving annoying to test but it's almost ready.
rbrunnerAsking around until get the full block?
hycI'm reworking the randomx integration patch, it has gotten ugly with the added tweaks over the past few months
moneromoooI do not understand that question.
hycand currently the daemon mining support is broken
rbrunnerI mean what means "sync pruned blocks"
moneromoooYou asked for pruned blocks when you can, rather than full blocks that you'd then prune.
dEBRUYNEhyc: Could you define ugly?
rbrunnerAh, ok. Thans
rbrunnerThanks
hycdEBRUYNE: two different code paths for main blocks vs altchain blocks
rehrarhello, apologies for tardiness
hycought to be able to consolidate it back into 1
hycbut need to step back and rethink the overall structure
moneromooo(it's to save network bandwidth btw, it doesn't save more db size)
rbrunnerHow does reworking of RandomX PR look in regard to the calendar?
hycwell, it always takes 2-3 days to test and verify that the network is behaving
jtgrassiefwiw hyc, current state of pr seems to be working now
rehrarIt's September, do we have a hard fork date?
hycjtgrassie: we haven't crossed a nother epoch boundary yet, I don't think
jtgrassie^ good point
moneromoooNo fork date yet.
rehrarWe were thinking Octoberish though, no?
moneromoooWe were.
hycI would assume mid-October
hycwe ought to be nailing that date down
rehrarso in theory there is a hypothetical freeze coming mid September? :D
moneromoooThough the randomx code being still changed makes me nervous about mid october.
tevadorhyc: for testing, you could shorten the epoch to 128 blocks, then you can test it in 2 hours
rehrarfluffypony luigi1111 ArticMine smooth binaryFate ?
hyctevador: we did that when the PR was originally written. for some reason the current problems never showed up then.
dEBRUYNEhyc, tevador: The recent change were made after audit recommendations or?
hycsuccessfully mined millions of blocks with epoch=128
dEBRUYNERecent changes to RandomX
tevadorthat was with a private testnet perhaps?
hycdEBRUYNE: changes to RandomX itself are independent of the randomx integration patch
hyctevador: true
tevadordEBRUYNE: most changes were made based on audit recommendations
dEBRUYNEI see. I guess if we need more time we can always push it back to end of October, but most people and services are expecting a fork in October
rehrardEBRUYNE: are they though?
moneromoooNobody expects a monero fork.
hycwe could always release with daemon mining disabled, and fix it up in a point release if we need to
hycsince xmrig is already available
rehrarI realize we don't have to go Verge vaporware extreme where we push back indefinitely, but I think people are used to some "delays" if it means code that works from the get go (hopefully)
rehrarespecially if it's just a couple of weeks
hycbut I'm pretty sure I can get a new patch ready in the next couple days
tevadorhyc: I think you could change the testnet epoch to 128 blocks even for public testing
rbrunnerNo mining in daemon would make me nervous, I have to admit
hyctevador: yeah I guess we can try that.
endogenichello all
rehrardsc_ or selsta here also?
hycspeeding up test verification to 2 hours would certainly help
endogenici may be in and out
selstayes
rbrunnerWould also be a minor PR defeat, so to say, after telling everyboding about restoring everybody's capacity to mine
hycmy cat just died
endogenic:(
hyc^ joke
endogenico
rehrarI think launching with daemon mining is pretty crucial. Shows we are prepared and not reliant on just one software for it, no?
dEBRUYNEhyc: Oh, then we should still have plenty of time
dEBRUYNEThere's like 6 weeks left until mid october
hycok then should be no problem
dEBRUYNErbrunner: Yes I tend to agree. I'd prefer to release v0.15.0.0 with full functionality
rehrarselsta: what's the state of the GUI as we march toward this fork?
tevadoryeah, we should make sure daemon mining is working, especially since there is not much difference in hashrate between xmrig and monerod
dEBRUYNEWith respect to branching, I guess we just keep merging stuff into master until the RandomX pull request is ready?
moneromoooThere's certainly more stuff to be merged atm.
rbrunnerPragmatic approach :)
selstaxiphon added simple mode public node discovery without a centralized service, I did some redesigned the balance card, dsc is working on i2p
hycmakes sense
jtgrassiewe shouldn't release a pow change and it being reliant on some third party miner
selstaalso small things
endogenicmoneromooo: hyc i just got it
hyclol
rehrarit seems like this is shaping up to be a standard Moenro fork :P
rehrarwe'll have our standard debriefing afterward with our similar standard complaints
rehraranything else currently being worked on?
tevadorhow long was the code freeze last time?
hycoh well. the integration PR was pushed in May. if more people had been testing it since then we could've found this earlier
rbrunnerLast time was quite rushed because of the "ASIC emergency"
hycas it is, we found the problems on testnet, so that at least served its purpose
rbrunnerAre the Wownero people running smoothly then? If yes, why?
hycthey aren't using daemon mining
rbrunnerOh
jtgrassieamuses me how the problem exists in loki and wow
rehrardoes anyone take Wownero seriously as a testing bed? Serious question.
hoochuThis problem is likely will not be detected unless someone mined 3 long altchains with epoch boundary in the middle privately and exposed them to hyc`s testnet
hoochu*would not be
sech1yes, mining issue only happens when there are long altchains across epoch boundary
hoochuI had another initial attempt when did that
moneromooo"long" means that one block on each side would not be enough to trigger ?
sech1not necessarily long, I guess they need to have different seed hashes
niocRandomX was released for wow will only daemon mining, tthere were no 3rd party miners at first
tevadormust be at least 64 blocks I think
sech1so 64 blocks is enough if split block is chosen carefully
hoochuI suppose enough but noone tried to expose them before daemon miner even started to mine epoch boundary block
moneromoooOK. I'll see if I can add tests for this.
hoochuThere was a race
rbrunnerThat sounds like an awfully special situation
sech1yes
hycyes, but one which any attacker can construct
hoochuAnd noone claims that he tested all possible special situations
sech1other than that, it works
hoochuBut it's better test all of them
rbrunnerMining ahead 64 mainnet blocks? Good luck for that attacker :)
rbrunnerBut I understand of course.
hoochurbrunner, you're wrong
sech1technically, you don't need 64 valid blocks to do it
hoochumine 2 blocks before mainnet and expose them immediately
sech1just 64 blocks to trigged node to verify them
sech1and boom
sech1it's broken
hycanyway we don't need to occupy the rest of the meeting with this
hycdiscussion in -pow
rehrarkinda fascinating though
rehrarare there any questions about specific issues or PRs?
rbrunnerRemember when PoW algorithms were easy and simple …
rehrarno core team seems to be here though :/
hycrbrunner: I suppose we'd have similar problems with any PoW scheme that references previous blocks
hoochuthis problem exist due to complex dependencies in monerod and lack of people to know all of them to write correct code but not local small changes
hoochu*that know all of them
moneromoootewinget: if you're still working on loki: ^
rehraris vtnerd here?
rehrarI think he said in one of the previous meetings that his networking stuff will probably not be ready in time, correct?
dEBRUYNE<rehrar> so in theory there is a hypothetical freeze coming mid September? :D <= I guess branching is technically a freeze right? Because typically only fixes go into the branch
rehrarthough it doesn't need a hard fork for his stuff so it doesn't matter
dEBRUYNErehrar: Yeah I think he said the dandellion++ stuff would not be ready in time
iDunkvtnerd__: can you take a look please https://paste.debian.net/hidden/bccdc3a2/
dEBRUYNEHis white noise PR has been merged though
iDunkThat's a MacOS depends build with Clang 3.7.1
rehrarwas there anything else that needed discussing?
rehraralright, so it looks like we can call it here for the meeting
rehrardiscussion can obviously continue afterwards on various topics
rehrarI'll try to ping core team peeps to be present for next meeting since we're drawing very close to a fork
Sources and notes
- Meeting log: Overview and Logs for the Dev Meeting Held on 2019-09-08, 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.