All Core Devs - Testing #93

Transcript

Justin Traglia: Hey, Barnabas, how's it going?

Barnabas Busa: Early the birds. Are you just waiting on people?

Justin Traglia: Yep. Just waiting. I joined a few minutes early, because I always… Forget to join if it's last minute.

Barnabas Busa: Hmm.

Justin Traglia: For some reason, my calendar, like, sorry, in macOS, the little notification which tells me when the meeting is goes away exactly when the meeting starts. If I don't click on it before.

Mario Vega: Strange. Hello?

Justin Traglia: Hey, Mario. Hey, Pooja.

Mario Vega: Seems like everything's stacked in the EL and SL breakouts today.

Justin Traglia: Yeah. I assume that we're just gonna give a… like, talk about the definite for a bit, and jump into the breakouts.

Mario Vega: Yeah, that's… that's what I thought, too. Yeah, sounds good.

Justin Traglia: Yep, signed for me.

Pooja Ranjan: I added a last-minute agenda item. It's not big, just… reminder to authors to maybe approve the PR to move Glamsterdam and Meta EIP to review. So if that can be mentioned in the meeting, that would be nice.

Mario Vega: I'm gonna add it into the… discussions, list. Yeah, I'll touch the topic. Thank you.

Pooja Ranjan: Thank you.

Mario Vega: Hello, everyone. Let's just wait a couple more minutes before… Starting, so we can give some time to people to join.

Alright, let's get started. Welcome, everybody, to ACDT number 93. Today is August 24. We are gonna be focusing today on… Thank you, Justin, for sharing the agenda. We're gonna be focusing today on Glamsterdam DevNet 8, Proto-Danksharding. And there's a ton of, individual items for each of the EL and CL breakouts. So yeah, let's get started with, Glamsterdam DevNet 8. Do we have anyone, Stefan, or anyone else from the Panda Ops team who wants to chime in to give an update, or DevNet 8?

Barnabas Busa: Yep I'm not sure. I can give a quick update. So, we have very close to 100% participation now. Everybody seems to be proposing and voting correctly. What I've noticed is we have quite a bad incorrect, head votes on… on most of the epochs, in the past few epochs, we were hovering between 65% to 80%. Only adverts, so… That's something that we should probably take an explicit look. I'm suspecting that, Grandine is, pulling down quite a bit here, because they seem to be starting to vote on the, On each of the… slots at 4 seconds, not before 4 seconds, so pretty much every single attestation that they are voting on will come in late. I'm not sure if anyone from Grandine is here to comment on that.

Saulius Grigaitis: Yes, so… did you open? There is a flag, Which, disables this late, voting. Did you try that?

Barnabas Busa: while, like… After Gloas, voting after 4 seconds is technically… Nope. Really possible, or it shouldn't be possible. Why do we have to have a flag for this?

Saulius Grigaitis: It's a bit of a historical thing. So, well, let's, let's discuss that offline, okay?

Barnabas Busa: Sure, okay.

Saulius Grigaitis: Okay.

Mario Vega: Yeah, Toni?

Toni Wahrstätter: Yeah, because Barnabas was saying that he observes, issues in the recent epochs, because I was just… an hour ago, I was, starting to test, some more malicious strategies, and it seems like one of the Nethermind Prysm nodes forked off, and also Besu Teku missed the slot, and another node. So, I posted it into the interrupt channel, would be great if clients could take a look.

Mario Vega: Cool, thank you. Barnabas, anything else, or anybody else regarding the 8? Stefan, we can barely hear you. There might be something wrong with your mic.

Stefan Starflinger: Maybe it's better known.

Mario Vega: Yeah, mature.

Stefan Starflinger: So I just, did a RPC kind of interoperability. review. I can share it here. I also added it to the agenda. Because there are still quite some inconsistencies, especially with things like, ETH, estimate gas. It would be good if clients could have a look to see that the behavior of this is correct. Because there are quite a few, inconsistencies there. And there's also some other, less important things, as well in the report.

Mario Vega: Thank you. I've added every single item that you have into the breakouts, because I thought it was, like. Better to discuss them separately there. So yeah. the Beacon API items have been added to the EL breakout, and the execution APIs to the EL breakout, so we can go deeper in there. Okay, thank you. Alright, Barnabas or Stefan, is that it for Donate 8? Do we have anything else? Okay, any… anything else anybody wants to highlight from Devon Elite? From the client teams, or… That… that is important to highlight today?

Barnabas Busa: I would just like to highlight that I think this estimate gas is gonna be very heavily used, at least for the next few months. Just because all the repricing, would be very, very good if all clients were properly estimating the gas for transactions. Because I feel like lots of wild Devs will… run the estimate guess instead of, Hard coding, or maybe they will just work with a new value, but… I think lots of people will have to use estimated gas for determining what gas values they should be using for the next transactions. Yeah, like, Chase, I think it would be very high value to update all those tests for Glamsterdam.

Mario Vega: Alright, And Chase, do you want to touch that topic here, or should we take it to the EL breakout?

Chase Wright: I think we can just talk about it in the breakout, it's fine.

Mario Vega: Okay, sounds good. All right. Yeah, let's move on then, Alright, so non-DevNet 8 discussions left for today. General CL and EL would be Glanced Meta EIP discussion that was brought up by Pooja. Pooja, would you like to, chime in on this?

Pooja Ranjan: Yeah, sure. Thank you, Mario. Just flagging this for authors' attention, keeping the meta EIPs up-to-date is important because it serves the canonical source of upgrade information. Right now, we have already received approval from Editor, and this is up-to-date with the latest masters, so I would really appreciate authors' attention. The PR number is 11855 on EIP repository. Let's move it to review. Thank you.

Mario Vega: Alright, sounds good. Thank you, Pooja. Alright, just before we jump into the breakouts, there was one last topic brought up by Toni. He's asking whether we are gonna have a DevNet 9, or if it's planned. Do we have an answer there, or should we discuss this?

Stefan Starflinger: I think it would make sense, since the transition, For this public testnet was a little bit bumpy. And there were, I think, requests in the last ACDC from Lido to do a more smooth fork transition so they can test it. But yeah, it's open for discussion.

Parithosh Jayanthi: Yeah, I think the main question is if the other DevNet is just going to be the non-finality network, or if it needs to be another network that's apart from non-finality. So, my proposal would be to just batch it in with the non-finality network. So, we can have, we can have the fork transition, check how things go, and only after the fork transition do the chaos.

Toni Wahrstätter: Why wouldn't we use DevNet 8 for the non-finality?

Parithosh Jayanthi: It's a public network, there's a lot of people we want to, it's a named public network, and we want people to start testing repricings. And it's just a different use case, I guess.

Barnabas Busa: Yeah, so ideally, we also want, client teams to merge in all their features into their trunk branches for .9. So that's what we're basically waiting for at this point, I would say. So that, most of the core changes are… Kind of finalized.

Mario Vega: Alright, is it time to discuss a date today, or should we wait another week?

Parithosh Jayanthi: The main blocker for us is all the clients having trunk branches ready. Is there an update on when that's going to be the case? I know that Teku and Lodestar are already ready, I think Nimbus as well, so it might just be Prysm and Lighthouse that we're waiting on.

Mario Vega: Anybody from Prysm or Lighthouse would you like to chime in?

Eitan Seri-Levi: No, I think… I think we could be ready in the next few days for Lighthouse.

Barnabas Busa: Could we actually try to set a deadline, for ACD, So that we can target, a launch of the next DevNet, by maybe early next week, Monday?

Parithosh Jayanthi: Yeah, we ideally also want to set a deadline and tell it to people like Lido, etc, because they have to set up infra on their side to track the DevNet. So it would be good to have a few days head, heads up.

Mario Vega: Alright, I don't think we heard from Prysm. Is everyone okay with the, Thursday deadline for this? Alright, if… okay, let's… let's do that, let's… let's go with the deadline for Thursday, and then we can… if there's any… Considerations, please raise them async as soon as possible. Otherwise, yeah, let's go with that. Okay. Do we have any other general topics, for today related to Devonnet 8 or Debnet 9 planning? Should we jump into the EIP records? Okay, if not, yeah, okay, let's, let's jump into the breakout. So, EL breakout, people remains in this call, and then the EL breakout, we can jump into the call that was, just linked by Justin. Alright.

Thanks, everyone.

Justin Traglia: Hi, y'all.

Mario Vega: Alright, should we get going? Yeah, first item in the agenda, let me just share it here, it was… Okay. It was shared by, Toni. This is, a specification for the Block Access List Getter. Semantics on the execution APs, number 8. 851. Toni, do you want to chime in on a…

Toni Wahrstätter: Yeah, yeah, not much to say about this PR. It's, it's a manfold, and I remember by the time it was only, like, the goal was only to fix some inaccuracies. Most clients have implemented it already, but it's great that we have it on their agenda to give people a heads up again, and also it would be great if we find someone who can merge it.

Chase Wright: So, I have merge rights, and I'm not opposed to merging it if everyone says it's good. Yeah. But, like, my only hesitation, the reason why I hadn't reviewed it yet, is because it deals with blobs, and like I said, Hive Chain hasn't been updated yet, so there's no way to actually, like. test these in HiveChain yet, and make sure all the clients are agreeing, so I hadn't done that yet. But if the specification is good, and we just want to merge it, I'm fine doing that.

Toni Wahrstätter: Okay, yeah, this is a good point. Should we maybe just post it into the… I can post it into the BlockHub Access's Discord again, and tag all the clients, and then we… Get there okay, like, async, and Then we can merge it and do the hive tests afterwards.

Chase Wright: Yeah, yeah, that's fine. If there's a Discord discussion and people are saying thumbs up or whatever, that's fine, or if clients want to just chime in on it and say, looks good to me, then I don't have any problem merging it.

Mario Vega: I think… I think we have all the clients here, so we can just make a decision. Rather, I think I would… I would say, does anybody have an opposition to this? And there's one comment from Karim, do we have tests to validate this one? I think that was already talked about, I don't think so. Chase, what's the… is… is HiveChain needs to be updated, right? So, but what's the… is there any other blocker, or should we just contact FedEx? To get this merch.

Chase Wright: the… Yeah, I mean, like I said, we can't write tests for it, because we… Because HiveChain isn't updated. So, normally, we would have some tests to go along with this, but I don't… It's not a mandatory requirement, so if we just want to merge it, we can.

Toni Wahrstätter: Yeah, I just… I just looked at the specific PR, and… most of the stuff is already merged anyways, and the PR itself, is only changing, like, small, small things, so I think we can merge the PR, because we have to do the tests at some point anyway, and most of the stuff is merged already.

Mario Vega: My client raises a good point. Do we have a PR specific for HiveChain, to update it? Is there a willing to… yeah.

Chase Wright: So, I do have APR, but it is in draft, because it's a very big PR that is kind of like a meta… MetaPR, and it's this. But, you know, I will fully admit that, normally this work is done by Felix, and Felix has been out, and so I'm… I'm doing my best, but I don't… I don't know all the moving pieces, so…

Mario Vega: Alright, let's contact Felix. Oh, he's out on vacation, right? Or… Barnabas, I don't know. I don't have the answer to that question. Do we have.

Chase Wright: Yeah, and there's also… there's also this PR, which would help fix some of the currently failing tests. It's not specifically related to Glamsterdam, but it is… A problem that needs to get fixed.

Mario Vega: Okay, that second one, I can help out to merge. That's pretty easy. The first one, yes, we need Felix or someone with knowledge of HiveChain to merge this. And, like, client, do you… are you referring to… They have CHINPR?

Lightclient (Matt Garnett): First one. JMPR, sorry, I meant to look at it last week or the week before when we were talking about it on the RPC call.

Chase Wright: Yeah, I know, lightclient, you commented on two of my PRs, and I responded, but I'd like to get clarification on how I can Continue.

Lightclient (Matt Garnett): Yeah, let me take a look at it and see.

Chase Wright: Okay.

Mario Vega: Awesome. Alright. Yeah, okay, I think we can move on to the next one. Let me just share the link. This is the one, this was raised by Stefan. Stefan, do you want to chime in on this?

Stefan Starflinger: Yeah, just wanted to check in on it. It hasn't been merged yet, but I saw that clients started implementing it already. And I just wanted to see if there might be some feedback on it. It conflicts a little bit with what Chase has been doing, so either I would have to rebase on what he has, or he has to rebase on what I have, but I think it's very small. So it shouldn't be a blocker, and people have already started, implementing it, so I just wanted to… to bring that up, and get some awareness on it. I think NetherMind has It implemented fully, and ethrex, partially. And otherwise, I didn't look into it that much more, but maybe never mind, if, they have some opinions or something, that would be…

Nice to share.

Mario Vega: Alright, any opinions from Nethermind on this topic specifically?

Stefan Starflinger: Or in general is also good, because it would be great if we could get this also progressed and merged.

Chase Wright: So, from my perspective, this PR has the same problem as the other PR, where we don't have any tests to go with it, and we can't generate those tests because… hype chain. But, I mean, it… it… it aligns with the work that I was doing, and the other PR. And everything, so, like, I think it's fine, but until there are tests. I… it's hard to move forward.

Stefan Starflinger: Yeah, I mean, clients have implemented it, and, I think… Would it be possible to… Work on a fork or something, to not be blocked.

Mario Vega: Sorry, Stefan, what do you mean with working on a fork?

Stefan Starflinger: a fork of, high… To just get something running. So Chase could maybe fork, the Hive repo, and try to see if We could get some, things running before we merge it into the… Main branch.

Chase Wright: I mean, I can do a lot of that stuff, like, locally, but there's, like… so many interdependencies. Like, I'd have to fork Hive, and then also fork execution APIs, and also fork GoEthereum, and then somehow manage to put all of that together into… like… the actual hive run. Like, I can do it locally, but… That's tough.

Stefan Starflinger: Yeah, maybe after we can figure something out.

Mario Vega: Baras, would you like to,

Barnabas Busa: We're gonna just run these specific branches of Hive? Instead of needing to fork everything, we just run this specific branch. Wouldn't that be…

Chase Wright: I mean, yeah, theoretically, if we find someone with the rights to create branches on every repo, then we can create a Glamsterdam-focused Branch on all of the interdependent repos.

Barnabas Busa: Yeah, yeah, I mean, that's a non-issue. That's a non-issue. If that would work, let's just do that.

Mario Vega: Alright. Okay, so it seems like we can create the branches, and in the meantime, I hope the, HiveChain PR gets, resolved, soon, and then we can move on also with this. With this PR. Yes, Minister?

Spencer Taylor-Brown: Yeah, can I just quickly ask what the process is with the execution APIs, tests, and HiveChain? Like, my understanding now is that the process is… You make a change to the execution APIs repo, you add tests there, HiveChain then imports those tests. And runs them, so man… so… based on this, I feel like the test should pass before we actually merge the… execution APIs, PRs? Is that, like, the feeling?

Chase Wright: Yes, the… the holdup is that… the RPC compat… test… tester, it runs a specifically generated chain we call HiveChain, and HiveChain is generated through GoEthereum. And HiveChain itself needs to be updated to support all the Glamsterdam PRs, and it is not. That's the blocker.

Spencer Taylor-Brown: Sure, I guess, like, from my side, does it all that make sense to just… Focus on the blocker, and, like, if clients are… Like, able to… because if that's the main blocker to even write the test, then, like, yeah, it feels like extra work.

Mario Vega: Yeah, my client's gonna be reviewing that BR, so that blocker should go away, technically, when that happens. I just wanted to get the discussion for the other PRs, and maybe there's something else that is not blocked. On… on this list. Okay, I think it's gonna be the same story for the next one. Let me see, this is the other one that is on the list. Chase is also by you. Is it… does this one have the same issue?

Chase Wright: Yeah, and this is directly related to Stefan's, changes. So, he made a proposed change to standardize the method, and… But his change is… like, on top of Glamsterdam, and then this change is… It's basically the same thing, just with tests and stuff, but, but again, until… Until all the Hive stuff lands, I can't do… I can't merge it.

Mario Vega: Alright, is there anything that we can discuss today that is not blocked by this, or do you want to, Keep on discussing this. We still have another one, create access list, it's also that blocked. By… by the same issue? This is, 84854. Just go to here. I think that's also what you, Chase. Does this have the same issue as the other ones? Are we still blocked by.

Chase Wright: Actually, no, this one is only blocked because GoEthereum disagrees with every other client. And so, I created a PR to GoEthereum, it's linked in that, and it's… Here, I don't… I don't know, you know, it needs to get reviewed and either merged or… Tell me no, either way.

Mario Vega: Alright, can we discuss this then? So, what's the base PR? Sorry, the execution API PR? Description for this.

Chase Wright: Sorry, sorry, what? What's the question?

Mario Vega: Sorry, yeah, what's the, execution AP's PR, description for this? The et underscore create access list. What's the, what's the… what's the summary of the change, basically?

Chase Wright: Oh, the… basically, when clients are using Create… ETH Create Access List, if fields are not passed in, like gas price, max fee for gas, or whatever, GoEthereum is… using a default value that the other clients are not using. And so, basically, every other client returns one… one specific gas used, and then GoEthereum returns a different value, so…

Mario Vega: Okay, yeah, this seems like then it's only a GoEthereum issue. Is there anything left to discuss for this, or can we just merge it?

Chase Wright: Well, I can't merge it until GoEthereum fixes their stuff, because there's a dependency on Hive that GoEthereum has to be right in order for the test to be right.

Mario Vega: I see the problem, okay. Anybody from GetH can chime in on this? Alright, it seems like now. Spencer, yes, I… Would you like to chime in on this? I don't know if there's anything to discuss, but yeah, I think it's a blocker at the moment.

Spencer Taylor-Brown: Yeah, I guess, like, it feels like we need more ownership of… of, the HiveChain process, it's clear that, like, it's not… I, like, I'm… personally, I'm glad that, like. There's a blocking process for passing these tests, because… It means that we're not gonna have this mess of, I guess, API calls all doing different things. In the future? But yeah, I think… yeah, maybe one, like, one of us on the team can also take a look, but, Yeah, I think we need more, like, eyes on hygiene, that's, like. my main, feeling. And then also, if, There's a question of if GetH should be a dependency of HiveChain, and, like, I'm not sure what removing that dependency would look like, but, it might speed things up as well. Yeah. But these are open questions that I think we can discuss async.

Mario Vega: Okay, yeah, I don't think we're gonna resolve those dependencies here, I think we can just jump into the last topic of the agenda that is not blocked by HiveChain, in order to speed things up. Alright, and we have… the last topic is, Glamsterdam devnet 8, Kleinbox, there's a list that was raised by Pawel. In the… in the agenda, let me just link it here. And see, yeah. So Pawel, I think you're in the call, would you like to chime in where this found, and so on and so forth?

Paweł Bylica: Yes, hi guys, yeah, I basically started fuzzing, Glamsterdam's spec, This is the outcome. So, I don't think that we need to spend a lot of time, but just please fix this. So I can continue. I think some of them have been fixed, some of the… I tried to make, execution specs test as well for these. And some of them have been merged, so I think… this is not stalling so much. I think the progress is good, but still, there are some issues that… Will not be addressed so far.

Mario Vega: Alright, thank you. Are discussing any issues on the 8, or is this, like, a live issue, or are we… Is there any action to be taken, or there's only the need for tests?

Paweł Bylica: I'm pretty confident it can be reproduced on the life. network. So, yeah, there's a way to… To split the network. But I didn't try it. So I… I don't think this is… this is the approach, but yeah, I… I don't know, this is kind of the start. I kind of expect some of the issues to show up from fuzzing, and they did show up. I think the scope… It's pretty broad. This is… I mean, this is, like, the strong candidates, blob access list, and oil gas repricings. Right. Yeah, I think… Yeah, I didn't… I think, I might have… more information. over the next week or two, if they are fixed, I can resume this,

For now, these are blockers, like, partly blockers, because… just the father will hit this continuously, and it might hit some other… I mean, they might hide some… some other issues. But this is the first iteration, so… I'm not very worried, but if they will… I mean, more issues will show up every week. That's probably the problem we'll have to discuss.

Mario Vega: Absolutely. Thanks. I think we are gonna prioritize this for the execution specs, so we can, run them against all the clients and see if we can push fixes. As soon as possible. So yeah, Spencer or, do you want to take the lead on this, we can, we should make another release for this, to contain these issues, and then we have to test all the clients for this.

Spencer Taylor-Brown: Yeah, I think the only remaining PR we need to merge is the EIP-8037, bug that Pavel found? And then we could also include the, 2610… Dash removal, so… and that release. We could do that over the next day, I guess. Yep.

Mario Vega: Cool, sounds good. Alright, thank you so much, Pawel. We will, let any clients know if there's any issues with the, with the tests, once we have them. Cool. Anything, anything else, sorry?

Paweł Bylica: No, that's all, thanks.

Mario Vega: Awesome. Cool. I think that's everything for today's agenda for the EL breakout. Do we have any… Other topic that wants to be brought up by anyone?

Barnabas Busa: May I actually ask, what is the hard dependency… 5 on GoEthereum? Like, could it actually be just… Take it out.

Lightclient (Matt Garnett): Can't really be ta- taken out. and it'd still be, like, use, you know, using the same codebase, because it's so deeply integrated into, like, the Go libraries with Go Ethereum. the formats associated, like, with HypeChain, like, these are things that could live separately. You would just need to… you know, if you want to do it with eels, you would have to have the kind of, like, hive chain implementation in ELS, and you'd have to call the ELs you know, objects and… and libraries associated with it. But I don't really see, like, there's a way to… There's not a very easy way to integrate HiveChain, kind of as it is, and just, like, swap out with, you know, go ethereum with ELs.

Mario Vega: So, yeah, I think the problem is that ELS really does not have an RPC at point, and really would not be able to respond to the… To the requirements of high chain at the moment. I mean, we can create some… we can… we could create, in theory, something in yields. It would have the advantage of, like. Once we update the spec for every single new EIP, we'll automatically have the RPC expectations. Of the responses, but I don't think it's gonna happen very quickly. There's a lot of things that we have to implement. It's not easy.

Lightclient (Matt Garnett): Yeah, that's true.

Mario Vega: I think, I guess, is this, If this is such a blocker, I think we can start, like, allocating some time into trying to make yields, also. make the RPC responses, but I don't think… I mean, it's not gonna solve our immediate problem anyway, right? It's something that we have to resolve for TemNet 8. And this is definitely not gonna be in time. Maybe for the next work, it's something that we can aim for. If that's okay.

Barnabas Busa: I think it would be very valuable. Even if it's not for this, Fork, but for the future forks. Coming.

Mario Vega: Yeah, absolutely. Okay. Then we can start looking into prioritizing that in yields. For the time being, I feel like we still should resolve HiveChain, just to get this test out.

Lightclient (Matt Garnett): Yeah, I mean, we're not gonna change this before this fork rolls out. So we just need to… Need to get it updated.

Mario Vega: Alright. I think the next steps are clear, or do we have to discuss anything else? Yeah, if not, okay, is there any other topics that anybody wants to raise today? Should we wrap up? Go in 3… 2… 1… Okay. Thanks, everyone. See you next week. Or Thursday.

Toni Wahrstätter: Thank you. Bye, buddy.

Mario Vega: Right.

Marcos A. Maceo: You, bye.

Transcript adapted from ethereum/pm, CC BY-SA 3.0.