Dencun Interop Testing #36

Transcript

[Music] that's [Applause]

welcome everyone to our 36th uh Denon testing call um I guess the two main things I had uh on the agenda today were just uh any updates on Def 12 for many of the clients or testing teams and then um chatting about uh getting the gordi shadow Fork up if there's anything uh there and if there's anything else anyone wants to discuss we probably have plenty of extra time to do so um yeah I guess to start uh anyone want to share some updates or uh yeah thoughts about how Dev 12 is going

yeah I can begin so 12 still doesn't have prison on but other than that uh we've been quite stable so we did some blob testing last weekend uh maybe par can discuss a bit more about that and we tested some deposits and exits and the previously mentioned nimus Bo has been resolved last week uh we were talking about uh nimus Exit book that uh since has been patched up and uh we tested that it the fix did work indeed nice uh yeah by blob thing Bas do you mean the blobber or what yeah um Mario's

on the calling can of course go deeper but um yeah we use his blobber tool to test equivocations and we had slashings as expected on the network uh I think there were about five SL validators and then we turned it off to wait until prism's online to slash some more but the setup exists uh M's been deployed on all nodes um every client pair that we've had on there has so basically abcl has at least produced One Mev block um which is perfect uh the only one that was missing was I think loar but gajinder pointed out that we haven't added the Builder flag on load star's validator client so once we do that I

presume it would be fine um we'll be doing that soon but uh yeah otherwise we also ran the regular check that every single node has produced at least one block with a blob and every single combo has done so every El as well as CL has in fact done that yes be a builder API with blobs as well um in case uh it's hard to reproduce that command um I will try and just um paste the result yeah it's not going to be very readable but you can see we we basically have um Dora which is the Explorer and that's our vers it's just storing everything to a

database anyway so Philip just wrote a nice SQL query for us to get a nice graph as to which combination has proposed what and it also connects to the relay API so we can easily figure out which blocks up proposed VI Mev is the blob rate simply the amount I guess for sorry another way toer this question for the blob rate um were we trying to get to one like is the ideal that no okay I I don't think so I think it's just a indication of how many were produced it could very well be that there was no possible blob to exactly there was just nothing valid that could be included got it the main reason we just have that number numbers in case

there was one node with like 01 and the rest of them are like really high then that's indicative of an issue um which in this case is true with loadstar ethereum GS because the node is down like half the time yeah very cool um yeah any other observations comments thoughts on the devet yeah it's it's really cool to see

that we have the entire pipeline working um I guess yeah next up is the shadow Fork then um I don't know Perry Barnabas do D you want to give us an update on uh on on how that's being set up next up would be onboarding prism first and then if all the blobbing uh test pass with prism then we can discuss the girly Shadow Fork I think you'd want to it would be just yeah we would want to wait with prism uh otherwise we would might need to repeat the shadow Fork which would be very yeah the idea was that we wanted to do a really big Shadow for so it would

make sense to wait for prison to do that and in the meantime just to make sure our tooling is okay we've been doing work shadow forx um you've been doing what what type of shadow veral shadow Forks like on the work testnets and I mean we've been finding work bugs but that's a different story um but the plus side for us is that our tooling is really nice got it and would there be value in doing a smaller Gordy Shadow Fork while we wait for prism or um do you think that we're basically getting all the information we like a small Shadow Fork won't give us more than what we get on the dev Nets already the problem is that 40% of the network if not more is prism so prism is

really a right right right a majority client so if if it would be one of the minority clients with like a few percent uh then I think we should be good to go but uh so what does it do actually so if if 40 % of uh Gordy right now is prism when we Shadow Fork the we start a new Beacon chain so we would just choose right right right yeah we would use a new split um the main reason it won't or I think that it won't teach us too much is that um 4844 is mainly a networking related update so having like a small there's not much

overhead in terms of producing a block got it yeah okay um and how uh is there someone from Prism on call uh I'm here if you like some updates um yeah yeah yes we we've uh a a big blocker for us is that we um we had kind of like a work in progress uh blob storage solution where we were putting our blobs in um our database and we wanted to make sure that we were testing our like final file system storage solution um so that's that's complete um we're still still working on pruning but that doesn't matter for this kind of time scale um and uh We've made some big changes to um our verification code um

which we've implemented in gossip but um not in RPC so that's kind of the last pieces missing um and um I'm also working with uh David and Justin on security um they're doing some extra code review got it and I know I I always feel shitty asking for timelines um but I guess bar of us front R me in the chat I think we're pretty close I think you know we I'm sure we'll be done I shouldn't say I'm sure but I think we'll be done by before the next testing call um uh I think you know we'll check in kind of towards the end of this week um but yeah we're making a lot of progress I think we're close yeah and I guess the the the reason I ask is

um is there so so basically uh next week is pretty much the last week before everybody will go on holiday so like do if if uh prism is not like ready to test on a shadow Fork by next week do people feel comfortable pushing back to Shadow Fork to like early January or would we rather have a shadow Fork even with a different network make up before the holiday so that if we find bugs then um it might give us a couple or some of the client teams a couple more weeks to fix stuff um yeah any thoughts on that I think it depends on prism at this

point yeah yeah if you can get prism on boarded uh and working by like let's say Thursday then probably we can do a shadow Fork next week okay yeah and girly notes shouldn't take too long to sync um I think we still need to tune like our Aon sync parameters sometimes it takes a bit long um but if we can trigger something maybe before Christmas weekend or something uh Bas and I should still be around afterwards to do it between Christmas and New Years okay yeah we can we can get uh devops a build sooner if you want to just start

trying it um we're still testing and um you know we know there's problems with uh RPC but um but gossip um should be working um so we can we can give you a binary to try we just need a branch if you give me a a branch I can build it off of that okay yeah we can the gossip changes emerge to develop um so um yeah just I don't want you to waste time on something that that could blow up because we haven't done a ton of testing yet um on on the Gosset path on the yeah we don't we won't uh probably immediately do it on devet 12 anyway we just toss it through kosis and likely Hive and see um see what breaks

yeah we'll coordinate in our uh Discord Channel awesome um yeah anything else on the Gordy Shadow Fork um if not I guess have other clans already started merging stuff into Master because it would be nice at um we at a place where we can cut releases from master and Jan yeah okay might house getting

there I think thatu just uh merge there changes also just get uh message in intro oh yeah I just uh deployed the teu changes on te get one Eno if you want to have a look logs look fine um but if you s check it then I can deploy to everything yeah I'll do I do the check now and I guess

is does anyone think they won't be in a spot they have their changes merg the master by January where they see like at least another month before this happens okay any other testing topics or oh uh yeah that is a PR like client posted uh should we merge

this I guess m is in here yeah but I mean I can ask him offline I don't know if other clients have looked at this in general it looks like most of the people who've been involved with it have kind of approved and so it would like to merge it today or tomorrow but yeah just if anyone want to call a second to take a look and Sanity check I would be appreciated yeah and maybe it's wor of just pin Al cordev and if nobody objects in the next day or so we just merge it yeah I'll check with me and do that sweet this is mostly just a cosmetic

change right yeah it's really just clarifying the actual like precedent of errors so now there's you have a better understanding of did the E accept that um that for chice updated or not okay okay see Mario will review anything else people want to bring up there's a a small change on execution

spec test that uh we working on um mainly this PR um this is um we're going to include the state test format on the execution spec test because right now we have only blockchain test format um so we are uh yeah basically we're we're we'll be able to fill all the test us say test from now on also as as well as blockchain test uh I just want to give everyone a heads up uh if if there is any C that doesn't support this status format I think most of them should but if uh if any any anyone does not just let me know and probably we will make a

release soon in the next couple of weeks with all the estatus format uh test filed um so yeah we we'll let you know and we will try to run with all the clients uh whenever possible when when we are ready sorry awesome uh it's a bit more of just like an execution spec test question but what aside from the state tests which uh parts of ethereum tests does it not cover so we are missing state test format and also the trans transaction test format got it uh those two but we're working on satas I think is the

most important and pressing one uh because we probably are missing coverage there got so yeah we need to uh get this ready as soon as possible sweet and then after this probably the next step should be just to start porting over tests from tests to have they full coverage of yeah but this is important mainly because it this increases canun coverage so I think this is the most pressing issue right anything else anyone wants to bring

up if there's nothing else uh the last quick thing I had so the next of these call is is supposed to be two weeks from now which would be exactly on Christmas Day I suspect most people will be out that day um one idea so I guess you know we can either uh either just skip it um but one other idea that uh was still Tempa uh I was talking with Danny about this I don't know if he was fully on board but basically um this the ACD during Christmas week on the 28th I don't think either me or Danny will be around um to run it so if people want to have a

testing call on that week it might be good to just merge it uh merge it together and have like an optional call on the Thursday for anyone who's around and wants to catch up um otherwise if people would rather not have a call at all that week we can just straight up cancel it um I think uh Danny and I will be able to handle acds on the week of the TW the 18th and of the 1 um yeah anyone have thoughts preferences on that I guess does anyone want to keep the call on the 28 feel like there's value in having

one otherwise we can just default the canceling for that oh Barnabas Barnabas wants the call so Barnabas might be here running the call i' be here alone I guess okay so barard this will be here giving you P up to anyone who shows up on 28th um I'm on 25th oh 25th I don't think we should do 25th I think we should okay I think we should just if we want to have something at all that week I would just do Thursday 14 UTC and have that be a testing call and not like an aord devs so that anyone who wants to talk testing has a space to do it we can also decide this we can like delete it for now and

decide it on the week of if we want it yeah maybe we just decide closer to the date depending on what's there to test cool so yeah so let's uh I'll put it on the agenda for uh for award devs not this week but next week um and we can decide if we want to have one um but for sure I would not have it on the 25th I think that's probably the worst possible time um sweet anything else okay well thanks everyone um talk to you

all this Thursday on ACD cool bye bye goodbye bye