Ethereum Foundation Core Devs, 12 Oct 2018

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Ethereum Core Devs Meeting #48 [10/12/18]

Description

Agenda: https://github.com/ethereum/pm/issues/59

A

Alright welcome to the call everybody we're gonna go ahead and get started on the core devs meeting number 48. This one should be a short one, because no one added anything to the agenda, so we're just gonna mostly go through updates and then talk a little bit more about Constantinople. So to start, do we have a Dimitri or Martin or anyone else who can speak to testing right now.

B

Yeah I guess I can talk a bit about it. So yeah the tests are not just stabilized, but there has been. There is still activity in generating the tests, and earlier earlier today we found a consensus issue in the CPP ethereum engine, which has caused some bad tests and that has been fixed. So we're waiting for those to be regenerated again, I think acts. I can't talk, I can't really say how wide the coverage is. So the test Dmitri can talk about that. It becomes when he I'm here, yeah sorry yeah. You can go out.

C

So the topic is a test right. Yes, so yeah actually doing good I am receiving reports of failing tests from different developers in the chat and we fixing those issues. Like Martin said today we fixed one issue in CPP and just already regenerated and being merged into a test developed. Well, it is actually good I hope we find all the consensus this year during this test period, but still speaking about tests which are still in development. It is Xcode crash test and a couple more tests, around storage changes and blockchain transition tests and yeah.

D

C

Think that's it difficulty formula check tests are also being merged last week, so you could try difficulty for a long on folder called basic tests. There's difficulty Constantinople Don Jason, those are Auto generated test for difficulty. Formula tells yeah I! Think that's it for now.

B

Yeah I can also mention that the first testing framework has been running for a couple of weeks and it's yeah. It's just done millions of tests, it's found some false positives and then it actually found a consensus issue, but it's one of those concerns issues which cannot be triggered on maintenance because it has to do with how non-existing pre-compile behaved. I guess, that's it all.

A

Right thanks, you too Dmitry. As far as the have you been getting a little more help with any of the testing, or do you still need somebody.

C

Know developing the test cases in addressed in adjacent formant, I didn't gotten your help. Jared was writing some tests, but she's only been focusing on working music, whatever she was working. This so I know to cover his boot requests and working on ex-con hash test alone. Also I'm Xing bugs in CBP client. There could fix.

A

Okay, so this is just a call out to any of the other teams again, if you have anybody who can help Demetri with some of these tests, especially I'm.

C

A

Them yeah go ahead: Dmitry yeah.

C

Actually, just today, I want one of my friend to reply to my message and she says that he has somebody who might be interesting in working with the serum test. So this evening, I'm gonna meet him and discuss this issue. Wonderful.

A

Cool, if you need any help with that, let me know but yeah I'm guessing your good.

A

So, let's see that's testing will go through kyon updates and, let's start with parody.

E

Yeah so far, the only we use is that we did a constant in open releases today, so we are prepared for ops and what is a Coburn hard fork? That's coming up this month, I think last time, I shake. It was Rob Smith this weekend right.

A

Yes, Sunday right has anyone else been looking at that.

E

So he has that sets it from parity. So if anyone wants to test too hard for to Constantinople changes on the parity client, please upgrade to version 207 or two one. Thank you. Thanks.

A

So much gift I.

F

Guess is the same as parity with the release on Monday, which I think on Monday, maybe on Tuesday morning, I, don't remember which enabled all the IPS and order basically the wraps and heart Fork. Apart from that, nothing spectacular to report.

A

G

Nothing super excited. We just been working more and more on sync performance to get our the amount of time it takes to sync the node down to a reasonable level, I'm still making pretty big leaps and cutting that performance number down, but still a lot left to do.

A

Okay, a theory MJ or harmony. We.

H

Have published these 30 minutes ago with enables Constantinople on Robson.

H

There is a problem with tests, so yeah at the moment it here MJ doesn't pass all this because there are some test issues, but I hope there won't be any consensus issues, but what worry that it could be covered? The truth could be fine by my dad s yeah. So that's it.

H

A

Thank you, a less.

C

So we released version wine 1.4 and the this one was stabilizing for some time, and this is most in an effort to have stable releases from time to time. So this one is not fully Constantinople compatible, but the de master master brand is, and we have one node running C++ code on Robson and and I think we switched to parties network monitor. So it should be there on the list. So I think that's the only one that will be going through the Europe's turn hard work.

A

Okay, thank you. Let's see, Nimbus.

I

J

Quick update you've been working continuing on sing continuing on getting the general state tests going, making steady progress there on testing in general for for the beacon genie, if you're in 2.0, there's work being done on the test format there as well.

J

So if you're interested in in that just just jump into the beacon chain project on the theorem there's, there are issues open about that other than that. Looking into platform making sure we can compile on platforms such as arm and and others painted.

J

A

D

Yeah, the big update for us is that we have implemented all of the Constantinople in the eyepiece.

A

Awesome, that's great, let's see is turbo death here.

A

Okay, aetherium, je s, left a comment that says can't join the call. So the update is that we are making good progress and now have substantial parts of all e IPS implemented latest merging the s store implementation, so I'm somewhat confident that we can do a Constantinople ready release of the VM within the next one to two weeks. So that's exciting for them is mana here.

A

Hey Andrew, I think you're, unmuted and I. Don't know if you're on delay or if Oh, perfect, okay and Andrew you from mana. Yes,.

K

I am Oh wonderful,.

A

Thanks all free for letting me know if you could give us an update for your client.

K

Sure yeah we are, we've currently updated all the VIPs and we're just working on syncing Robson's. At the moment. We we started to focus on that last week, we're at about block 300,000 at the moment and main that we are at 1 million. Seven hundred thousand.

A

Some thanks for the update is nether mind here. I think I need to invite them to the call I, don't know if I really have yet I need to get in touch. What about X theorem.

A

Okay, other than II was, um and research have I missed any clients.

A

Okay, also, if you go to the core devs notes, there's a link for Constantinople progress and it looks like right now. Let me count but pretty much. Every client is completely compatible, except for Trinity aetherium J s and X theorem.

A

So that's really exciting and I'm not sure how many of those have the updated create to e I P specs, but I was seeing comments that most of them do at least all the major clients do. So that's that's all good.

A

Next we're gonna talk to e wasum.

L

Hey Hudson, mrs. Lane I can be was and unless there's anyone else, just adjusting the sound cool yeah, so I mean the main thing that we've been working on the past few days is actually benchmarking, which is obviously a very important part of this project. Right.

L

We want to understand what sort of performance improvements we're seeing both using the wasum engines like binary, inand wabbit, which are interpreters as well as a JIT engine called watham, just understanding what sort of performance improvements we're gonna see with with us and relative to EVM I haven't personally been working on that myself, so I'm not going to speak to the results of that, but I'm sure you'll hear more from us on that soon we've been continuing to work on tests and stuff, that's going pretty well, we've got a stable test net running and we're still planning on that being public as of Devcon.

L

So it's kind of documentation around that as well. Hallo I think you're here did you want to add anything Pavel on the eve EMC work you've been doing.

C

Yeah I mean so I'm I'm close to release a VNC version. Six and I have to update, go go bindings for that, but this some small decision you have to made. But besides that it should be stable for some time so.

C

Yeah, mostly, that is planned version that I think we we don't want to modify anymore for for next month's.

A

Ok, great is that it line.

L

Yes, Thank You Hudson, alright,.

A

Thank you all for me was on the next is gonna, be research with I think just think Danny's, the only one in here from research, yeah, I.

M

Think the most interesting thing going on is Justin continuing for with this PDF feasibility study and an effort to create strong one by one by school, randomness file coin is has agreed to split the cost of the next next chunk of this feasibility study, which involves actually designing some circuits and stuff because they are interested in the same reason.

M

Hopefully in the next couple months, will have like a really clear idea as to whether this is a feasible path forward for other random beacon to know that ran Dow actually serves our purposes in most cases and likely will be moved forward with initial iterations and that the VBF can be layered on as a hardening technique in the future.

M

If we choose then cleaning up and also adding a lot of features and components, the eqo spec there's a lot of development going on in that right now, I wanted to just again make sure I know that there's kind of two parallel efforts going on right now, but the people in this call and the people on your teams have a lot of incredibly valuable insights and experience to add to this process.

M

So please, if you want to be more involved, there are a number of like you used to aspect Cerises like needs two aspects: rico ETOP Emmys 2o tests where these things are beginning to form. Please take a look and please, if you want to, if you want to get the lay of the land, please don't hesitate to reach out to me to set up a call to talk about.

M

What's going on and maybe figure out where you can get involved, that's that I know there's a ton going on over here in terms of getting ready for those Forks, so maybe maybe after comms be a good time to pop them or drink.

M

That's it for us. Thank you. Thanks.

A

Danny and I think that's it for updates, what'd I miss any clients, did I miss any researchers or anybody pop in at the very end.

A

Okay, the next thing is the Constantinople and Robson hard fork, so we'll start with rob stone, that's gonna happen, I'm thinking around Sunday or Monday from the last time. Someone ran that estimation. So does anyone have like a new estimation by chance, if not we'll just kind of say it's around Sunday or Monday.

A

Yeah, so we're thinking it's around then and I think we're ready for that. We'll just have some people online as it gets closer hopefully and if anything bad happens, we'll all get together right. We usually don't have a call for something like a test net, but that might be a good thing to have.

A

Let's see we'll also give an update on Prague was just suggested by Martin, so we'll do that after we talked about Constantinople so with Constantinople, we are still doing tests on that. It's definitely not going to happen before Devcon. As we've said, the last few meetings it'll happen after DEFCON we're not exactly sure in about two weeks. The first groups of people are gonna, be arriving in Prague for Devcon, so I think we should skip I think we should skip the next meeting and just have a meeting about a month from now or for I.

A

Guess four Fridays from now. So just skipping one does anyone think that's a bad idea.

A

Cool all right, we'll do that then, and we'll have more on Constantinople at that time until and by that time, I think pretty much all the clients will be mostly implemented. It sounds like from everyone giving their estimations.

A

Alright, anybody else have comments on that. Yeah.

B

So do we preliminary do we have an idea, we've said after Constantinople, sorry after dokkan and right now, everything seems implemented in our clients and we're going with test nuts in a couple of days. So if we find no consensus errors in the testing room, if I no consensus on the test net, what would what should be aim for with focus on Tonopah? That's my question. That's.

A

A good question: anybody have any thoughts on that I'll pull the calendar.

F

Well, my recommendation would be that since December is kind of questionable month, so we don't know who is working. What is working holidays and whatnot I would say that we either. If everything looks good on Robson, we either push it maybe mid November, but if we, if we don't push it in November and I, don't think it's realistic to push it in this I would vote for mid-november or end of November I.

E

Would second set I think mid know the mouse might be a bit rushed since we I guess we have our next deaf coil mid-november. So I don't really see this happening, but if we want to do is just this year and maybe knit or maybe a November I.

G

Would second towards the end of November, but that isn't based on client stuff, just not doing it immediately after DEFCON.

A

Yeah that sounds good end of November and how many weeks do we need in advance of putting that block number in? Would we how many do we usually do I.

G

Don't know what historically we've done, but I feel like a month of like time between the the time that we set the block number and the time when it actually happens is a good amount of time to give people plenty of time to do. Updates.

E

Yeah makes sense months.

A

Okay, in that case, the next core dev call if we skip Def Con, is going to be on the 9th I think. Am I right wait? What's today the 12th? So if we skip the 26th, then the next court have calls on the 9th. We can decide the block number.

A

hmm We can decide it as early. As you know, the 25th 26th. We can decide it as late as just the next core dev meeting pick a number and just decide on that next chord to have meeting what the number is and if we're doing it this year, if we run into problems, we would push it off until January, obviously to avoid the holidays in December, so yeah.

G

A

G

It's not a it's, not a block number, but in terms of dates. I I would target something like November, 27th, 28th or 29th, which I think targeting later means. It'll happen. I wanted to slightly earlier days, depending on how far out we we target because of well how block times are not perfect.

A

So if we, if we do, you think it's too late to make this plan on the 9th or should we make it like literally immediately after Devcon, like in chat, I.

G

Mean I think we can get a block number announced for those dates and if, on the all core devs call on the 9th people are not confident about that, then we'll be scrapping it for 2018 and pushing it to January 2019 so like we can. Somebody can do the numbers and and estimate a block number for one of those dates. We can announce it outside of this call. Unless anybody has any objections to that, and then we can confirm on the 9th what.

E

About having a quick, Constantinople call in like 7 days just evaluating the fork on Robson, if that's possible and really keeping it short and saying this looks like it's running good and announced like a preliminary I, don't know block number and if something goes wrong during DEFCON or the subsequent reads weeks. We can just pull this number and say: let's do January or you think.

B

M

B

I won't have I mean, do we want to repeat the P Asko oblast fork where we put a number in the compiled clients, release it and have to do patch releases and roebucks or not? Do that and risk having splits done away or should we have some kind of a signaling mechanism and if so, how.

G

How about dance avoid pushing those updates until, after the all core devs meeting on the nights, that whatever pull requests, that is, can consid hanging until until that means yeah.

B

So the problem is, if we push it on the 9th and it goes into the release on the 12th and then there's a consensus issue on the 18th. There's still 11 days left to the fork, and then we have to decide. Okay, yeah postpone or not fallen or whatever yeah.

F

I guess this is the discussion we had at last, a song where we were talking, whether to implement bail out mechanism for hard Forks. How.

G

About we sign this and discuss this like some potential solutions to this, and we can see if we can set on something by the meeting next Friday.

A

Yeah that'd be a good idea, but I think that just to put put some items on the table before we discuss it, the main one I remember that was sounded like the most feasible is having a smart contract that pings, basically the clients can ping and see if the pork is still on. I guess is that kind of the main idea very.

F

C

Just got an interesting idea for the test if we clone an existing main network, and I ran a test cluster from a certain block right like from right now, every client is reaching the current win that transactions as if it was already won. Constantinople fork. That would be like another test network, but like a clone of current win network. What running on constantinople? So we could really repeat all the transactions as if it was already running on Constantinople.

A

Is that something we have the capability to do today or have done before, or is that just an idea just.

C

An idea I've been thinking right now at the moment, and it came up to me so.

B

I mean it could probably be done, but the only thing we'd really get coverage of is the nut test or gas changes, because no one would be using the shifting stuff and no one would be using create.

C

Yeah you're, right and block.

B

Delay doesn't bomb delay won't be noticeable for a while. Yes,.

A

Okay, so let's talk about signalling methods in the next week and yeah: let's have, let's have a meeting on the 19th. Why don't we and yeah? That sounds good to me.

A

So I think that's anything else about Constantinople. We want to talk about.

E

Just just one through say that I mean anyone. Who's interested can use the parity Robson dashboard, it's Rob's new dash stats, dot territory, dot IO and the WebSockets secret is the same as e stats. Dotnet.

A

Cool, thank you all right. So what's next Oh prog pal, so yeah anybody have an update on prog. Tau I know that they made some promises last time for some articles and for somes a little bit more tightening up of the spec and I. Don't haven't seen that but I don't know if anyone's been in communication with their team or not.

B

Yeah, so we have actually been in contact with our team me and Pavel, and also Andre from parity, and we have ironed out some kinks in the implementation. So right now, CPP. Well the e f-- minor stuff that power last encoding seems to match with the go implementation.

B

Now there are some problems with the EEP I, don't know Paul. Do you want to go into more details about that that problems with a protocol and stuff? Yes,.

C

So in terms of EEP itself, there are some big gaps in the spec, so it's I think it's practically impossible to implement this algorithm based on the EIP itself, mmm so I had to reach out for many source code. That was mostly done by the team and by the protocol team and still the final version was found somewhere in like one of the implementations and we discussed the differences.

C

So yes, as as Martin said, we we are on the same page in terms of going c++ c implementation, but definitely EAP is in bad shape and, and there are some details and some very small improvements, we can make.

C

But I think the most important part is some kind of we some kind of have to reference implementations, but there is no proper spec for it. So this is situation. I am not really I, don't really like, because it will make make the direction of where it goes, but I mean I, don't want to people to to to read, see or go code to figure out how to implement that.

C

But anyway, so the promise they did about improving EAP and answer questions mostly in didn't happen in public, but as much inside we have some contact with the people. So we couple of ask questions directly and you really get some answers so that helped me to finish the implementation that address produce the same results for the basic crisis and it was it was finished yesterday. So the SkyTrain recent news.

B

Yeah so, and there are some changes and it's been made, because the original idea about the invitation is that it has an internal epoch of 50 blocks when it was translated into aetherium.

B

It uses the same epoch as it that is 30,000 blocks so there, which leads to some problems, which needs an update of the protocols between the miners. The get work function aside from that I think the general idea is to basically launch some kind of tested for it as soon as as soon as possible.

A

Cool that's interesting to hear so that's good that they're working together. So that sounds like the agenda. We want to pushes for the EIP to be updated to a very, very detailed specification, which sounds like a good idea to me. Is that something that we've communicated to them, and that is in the in the works.

C

Yes, we did so I think, as of today, they still plan to update the IP with what we discussed privately and like fill up the gaps that order.

A

Ok, great, thank you all for the update.

A

um Was there anything else anyone wanted to discuss? It sounds like we're going to be skipping. The next chord F column we're having a Constantinople call in a week in order to not mess with anyone. Elses dev calls we're not going to be like shifting any other meetings in response to having the Constantinople meeting next week. So Danny you can do a sigh of relief. There we're we're.

M

On Thursdays, oh.

A

That's right, worrying.

M

About our thing, okay,.

A

M

A

We won't shift anyway, just because that's confusing at least for right now, so yeah I'll make the agenda for the meeting in a week and I guess: I'll see you all there is there anything else.

A

Yay for short meetings: everybody have a good day or evening goodbye.