 So, where can we go on to the stable release buff? I think what my intention was to discuss about the topics, questions that are open to stable points releases. There's some questions that are obviously open for my part of you, and that is how we deal with current security updates that have happened in Saas, but are going to be open to the last points release, because they need changes to the installer, but of course there are a lot of other things. The second stable release manager, Martin, is also listening to this buff by a previous beaming in the works, and I hope that if he has questions or things to say, then I'll always be doing it for him. Yes, no. If you want to talk about this, I just want to discuss about today, the size of the kernel update in the installer, which I'll just note down. So, it's their work topics, this tells them to me, and if I'm too fast speaking this way. So, what topics? I hope anything that comes to your mind now, they want to speak about it. The only way to start this is obviously last topic. So, do you want to start this last one? Is there anything you'd like to talk about about the kernel team? So, what's the issue? I think the fans could do it better, but basically the installer is part of the series, and then if you change the kernel, and you have these new jobs, there's good reason to do so, but then we have to build the installer. That's it. Specifically, if it's an ABI game change, which is what happened here, so the model is no longer compatible. Well, you could even leave the old installer in stable, but that would mean there's no longer a source version for those units in stable, which is not something we want to do, because it would keep me out. And it could make a source keepser? Yeah, that's been suggested, but that game gives you name conflicts, so it's not an obvious solution either. You could just take some sort of packages and upload them as a special package, which is an entity. Yeah, but that means a moment of preparation that we currently have. Also, it has been discussed before, and basically the outcome of those discussions have been we should update the ABI installer when there's a ABI change in the kernel. Unfortunately, having an ABI change in the kernel, kernel UDEPs means that current ABI installer images may break, some of them, those that are not self-contained, but download kernel UDEPs from the net as part of the installation. So usable that we want about that. The main issue at this point in time is that this is the first time we're going to be doing this, so the infrastructure is not in place. And what we basically needed was some extra space in the ABI package migration set up on the Mirror Master to allow us to get kernel UDEPs into stable proposed updates. That should be done now. It is done now. I checked just what you mean. They actually are in stable proposed. Yes, they are. Check out the builders, I just looked at FTP Master. Sorry. Which means that we now could start to build the ABI installer using step-post updates and then we need to get the borders of the different architectures to also rebuild the kernel UDEPs for their architecture which is currently going to require some coordination but should be able to go down very quickly. I have all machines, except for Spark, which you have, and WebSense is getting there. So I know the contract goes there. Okay, that's great. Something ends when uploading packages of proposed updates right now because there are no packages going to be proposed updates at all. They are starting to execute and the basic idea is that the Stabilize team gets access to the queue and can upload packages for the check-in service but the upload for the check-in part was not done yet. I hope that the system starts to speak but until then we can upload but they won't get it. Do you want to say this is the first time you've done this? First time you've done what? That move via testing? Yeah. Because it is a Stabilize team. Oh, it's Stabilize. There's just a point of it. It doesn't mean you have to go via testing. We have not had to do it because previous kernel updates have been without ADI changes which means that the kernel UDEPs were still compatible enough to just ignore it and security issues that will fix were not that relevant for an installer. So we have to just skip that basically. To bring Sasha to out we said we're just ignoring all the kernel stuff and we'll do it later. So all the intensity for R3 are only big topics in the kernel that have changes plus installer So that basically is the situation. I don't foresee any major problems. We should be able to use to be able to build the installer using testing proposal dates. The only problem is that we won't be able to test the installer very much. We will basically only be able to do the builds we used ourselves. And after the release has happened and the UDEPs have migrated to stable proper and these have been built we can then run the check of the installation from CD. Basically as far as I understand but as far as I understand the standard is if we do the proposal dates we can still test it afterwards to make sure it's not delayed. And it's not so far that we have to install it like it today on the release. That's correct. And the whole installer is supposed to be still continue to work. So we can say maybe it's a proposal date so new installation will appear. It will speak or whatever and it's all supposed to be still works as a proposal date. We will have to communicate this very clearly to our users so I would suggest using that announce for this. I don't have access to that I have to remember for the what was that? I should see in here who can kind of push it through. I can cry. I really think we should use that announce for this. And I think it should be the very first part of the release update made. And it should also be a permanent procession on the website. So we have to make sure of that. I don't expect it to go actually with the installer breaking unexpectedly because of this because the kernel still is basically the same. The kernel itself has been tested and nothing major has changed that could be expected to impact the installer. So from that point of view I don't really expect problems. One thing we might know is that Moritz and I are working on another kernel update so we can parallel. It's not an API change but that means that we'll be yet another kernel image out there when this happens probably. Will that be ready before RC3 or is it targeted for after RC3? What's the date for RC3? The date is going to be pushed back because we have not had these kernel units to stay with Moritz updates. Who had anything else finalized today? Okay, so I think that means that we can use the current RC386 kernel unit system for post-properties to do the test build of the installer, see how it works and if the new kernel updates are pushed in then we should do a rebuild for all architecture in RC386 using those new ones so we can do the coordinates with the newest security updates. Sounds good to you? Okay. One thing I would think is for future API problems that we're running to it would be an idea to put an API number into the kernel module that you have so that they have API. So then how can we add going to auto-detect that if your CD is broken you can't install some kind of skeleton. That does happen. Okay, so it is clearly already told to the user when they try to do it? If they use an old network image for an old fully-based installation then they will be told the installer cannot find a kernel unit matching the running kernel do you want to continue you allowed to but you probably won't have enough with anything that's remotely useful? Yeah, I don't know how many users would understand that as much as you need to download a new CD or something like that. Can't find a kernel unit please check for new installer. Yeah. We could possibly do that. Of course not. Of course not. Yeah, but not for a single. I know we couldn't do it for such a big translation I think so. I was wondering. But it could be done immediately. Yeah. There is one other issue with the installer. Yeah. That's only relevant for alpha. If you take the base installer unit which is responsible for installing the kernel into the target system for alpha that has the ABI hardcoded into the code. So it will select only the current stable kernel. So with the update we have to change that to the new ABI so it will accidentally be able to find the kernel to be installed. And with PowerPC it didn't have an ABI and I did this update I added an ABI to make that change. That's another minor change in kernel batch which produces the UDAPs. But it's not a niche it's just the config file of kernel batch, PowerPC and these have the ABI changes everything like with the kernel landing. So that's just one of the building kernel types. So once you have this work you have the first update in your UDAPs and then you are building the installers that go into my hand which is set up by FTPmasters. Yeah, we have to first get all the UDAPs for all architectures interested in post updates. Then I will do or Jeremy will do a build of everything in store against tenant post updates and upload that to stable. Definitely needs by-hand processing then the outbuilder is still pinging for the remaining architectures which also will be by-hand processing and basically then you should be ready for derogies and moving everything to stable. Okay, now let's move that so this will be built for the UDAPs. Okay. Very good. Anything more on the alphas kernel stuff you know for when you have to install it here? Maybe I have a question would be current installers on the UDAP installers I was very frustrated trying to figure out how to put together a driver update this but my situation with the installer would not install on the machine at a different self-range than the current one so long ago and I couldn't figure it out and I searched and googled how to construct and the driver update the driver update there are means to insert drivers, modules into the installer and access drives that may not be accessible Well, good discussion on derogies I'm for such a bit difficult but actually yesterday with security and kernel and discussion on how we could do it better in the future perhaps if anyone of you wants to say something about it Yeah, so we had to propose yesterday that maybe we can do more edge we can do something like an edge and a half where we add an additional kernel about halfway between edge and edge plus one so that we can enable more hardware and do security support for both and he said he talked to the DIT and he was able to And who are you talking about now? I'm talking about first Sarge Santana though Our only requirement has been otherwise that the kernel team in the security team has to be willing to support such kernel Because otherwise it's just useless to even start thinking about it Yeah and I've been afraid to support two kernels like we're doing now with Sarge but the issue there is that two cores old and nobody's really using it but if we're doing like an edge an edge and a half where neither kernel is going to be as old as Sarge Santana I'm trying to say that but what about like what kind of a user like I drive with install and I find that it doesn't access the hardware because I mean I need a way to say okay provide a floppy or a cd when you shove in so it doesn't have any load either you do it or something like that prior to trying to do this system with install we do have support for loading drivers from floppy I guess that's not how look at this in two specific specific particular cases I think mainly for network in today's detection if no network in today's detection is detected then it will last for floppy I remember that you were saying for hardest but I don't think so I couldn't figure out how to create that possibly need to be out the one objection I have to doing a floppy base is that less and less machines have floppy disks how about a cdb you would need prior to switch cds you would need to be able to do it from usb you would need to be able to download it from the network maybe you would be able to do it from floppy if you have floppy so that makes for a very complicated code I don't know how we do it so far we haven't detected it I think we may do it after the current team has done the non-free and the firmware split because at that point we will need to do something with that I guess the best thing is that the vendor just presumably had to produce their own install or cd yes but at the moment the best option for vendors is to create custom images which for charge again is a pain because of the major changes between 268 kernel with which the storage install was shipped and the current kernels which have really caused major rewrites of the installer with regard to hardware protection and so on so for what it's worth the way I do a lot of stuff for the company I work for is I actually started with GI as is and I already know it's different to 268 driver updates AdWords for us to backport from a recent kernel back to 268 and a few exceptions and then I started with that and do the whole build cycle like runs with the ITM and you've kind of all built with install cd well I don't want to say I don't want to build every binary again I replace the bits that I need you have burned the new cds unfortunately I suggest that to my management and for some reason this is a good thing I mentioned I can push back out of it my management tells me that when I try to afford the drivers that enable them they don't want me to cut the new install or cd because they kind of were not in the install or cd business they only want to do driver update to me as that for that's a good thing to me that's a business decision of course and now I can push back and change that vision but we make sure that it's clear that our cd is different and every package in there has a different character of your name and kernel has an identifier that says it's up to my company so I guess we're going to come back to the IT with invalid information and that's a good report okay there's something that has to be doing very carefully especially the first class I couldn't do maybe I could approach my management and see that everyone is working my time to help develop some sort of driver update this type of business I think that would be worth discussing okay there have been several discussions and I'm going back to the progress unless you think we're going off topic I think when you say we're needed this this we can take this off we have to take it off somewhere else because if we have more things on the kernel as di we could discuss we should take some time for it I have one more issue that is kernel related there are currently two packages in the new queue one is PCMCIA 2.4 the other is the 2.4 S390 kernel they should really be in stable proposal updates but because those packages have not been updated since the Saturdays in unstable they have to go through unstable and trickle down to as I said in the new queue there are lots of packages from the new queue that were there basically what's happening is if you upload a security update or a proposed update and it takes this version it's the same as in testing it just gives a warning and also programs have to test the proposed updates now and Lisa so the least I can say is that because it's a warning it's totally in error and she said no which was then afterwards fixed but I don't make sure what needs to be done I say a target proposed update but if the same version is in such in testing or it's large in edge and unstable it's a target proposed update it might as well go through other things I'll check it but actually it's a changes file make sure that we get some other side to get back to you but it still means there should be an upload to a stable as well or what if we rebuild those packages from unstable does that work? yes basically I think for edge there should still be 2-4 channels but not so far I believe I don't think anybody has come back and said yes we will I'm willing to give all the security support but 2-4 so I think they're going to come I have suggested to the release members fairly recently to make a hard decision on that and start next moment because that will make things clear for everybody packages that's involved but still has 2-4 related packages to try for more deals I think I'll make sure that we have a decision before you all talk I hope there is the person here that's been most pushing to continue with them I think he's okay with them going away probably we should finish with Holger and perhaps you and you as well I'm checking we'll do a decision on that I think what we decided is that there's no reason to preclude users from running 2-4 and doing their own 2-4 maintenance we don't want to break that I think that's what we decided users should be able to work with 2-4 we had discussed this very briefly yesterday during the GI involved as well I think the remark at the end it was people can always download 2.4 from private headquarters and build it from there as long as it works so if there is no 2.4 kernel in Edge then we won't obviously have you been working on getting the box fixed in 2.6.8 in 2.6 kernels? of course because there must be a reason why you're still running 2.4 they work for you they work for me a lot so the installation doesn't work I don't know how to fix that because I haven't quite married down the many changes that I did I will do it by myself well, does that seem to work then? it is but it's something to work on I know but it's also something that it works the way I do it but if 2.4 is better in that respect the I can only support distribution support it's someone actually going to actively check if things do work if there are new devs basically we've had in store images with 2.6.14 to 6.0 for the past half year so if people haven't been trying that for their hardware you've said you wanted to be supported that users can download 2.4 themselves after I've told you 2.6 there is someone checking that this does work despite dependency despite the download and compile 2.2 kernels on SID and they will install and if you've got all that moving half the distribution you can build and run 2.2 kernels as long as you've got hardware support on SID the kernel back at the main KBKJ still supports 2.2 and I've tested it in mites so as of mites I wasn't really worried about low level stuff but it's just if there's assumptions and you know basically I think you're lying but I would say a typical QA task yes I think we should note that as our mission how can we test it but it's not so important either it's not really an issue because so far we have 2.4 supported in store and we have tested that you end up with working installations so that's not really important the only thing a user has to realize if he does it, your level won't run anymore so he will need some other form of hardware detection installed so he will need to I can discover something that should just be in the release but it's definitely somewhere just by the package we definitely need help for this release because it's going to be quite a complex story for 4 years I think about what you're thinking now well you can still build and compile and install GLC as in 2.4 is the minimum version so you're not going to be able to run any binary system and as time goes forward you'll be able to run pretty much nothing in fact I talked to a couple weeks ago join GLC Maintainers and me saying that the current upstream GLC has dropped to 2.4 support but when I test I was testing on a potato machine but it was built on a same machine I think it was built on a potato if you see the issue it's really something that's built on 2.4 support sooner or later I don't expect it to be built on edge but not only for extra support she put her timing in this concern do you really like the 4 edge but does it sound like we'll be able to if I remember correctly and the total cost of the 2 GLC Maintainers for 4.4 as I'm looking at it so we can keep it up but of course there will be a data which will just stop 2.4 support like I don't think that the support running won't see something falling on sit anymore yes so the main impact 2.4 will probably be 4.6K if we still have 2.4 this archive yes so they may have to try to develop some architectures or they will have to work very hard to prepare for remaining issues that are resolved I think all other architectures are about ready to make streets to 2.6 or have already done it the other main ones are the MIPSA MIPSA architectures and M6 8K and now I don't really care very much about M6 8K it's not a little discreet at any base no and S3 9 S3 9 is almost ready to switch oh ok it's not a little discreet but I hope that this changes then with this M6 8K I currently don't see it for edge but it's important for this whole anime for this whole they have to make the switch before release if they want to be an experiment yeah we will see so ok another thing that we need to discuss about stable points release or stable release management now of course on the large edge release we have to talk about this discussion on Thursday and make sure that we have a lot of discussion time so something going on about stable points release well are there any questions or concerns about the possible edge and a half type current update does anybody there will be a lot of technical issues to do later can you what do you mean by edge and a half well we're just calling what we mean is a stable kernel update that takes us from say 2616 to whatever is current half way between edge and edge plus one and I thought you were not going to hold anything but 2616 that's what we're discussing at the time release we won't but we might move off 2616 no in addition to 2616 add another one it depends on where 2616 is going if 2616 gets all the ground from backwards then you could just do a 2616 upgrade it's essentially it's an old kernel it's an update it's not done because as we have said it doesn't even mind about the AI compatibility which is something I'm not too happy about right if it's a quality yes it doesn't mind to existing current modules for new features that's something else as a security update security yes, definitely probably it's sealed by an audience I was like let's say something like the Ember still works even after the AI change which is but if you say maybe it's just changing the interface because it looks better that way but it's easier to backtrack drivers then it might not be the case but of course there's now not really something inside it it's just a case it might happen and I think we just need to let you the changes say yeah, we are there then we can say okay it's a change okay or not and then we can say do we need another kernel or not well I think the execution is more intuitive we would like to have 268 because there are teams that are scheduled going forward 17 and 18 that will make it possible to deploy a real and really hardcore security environment multi-level security teams so there would be at least one constituency that would like to have support for the latest kernel out there from the Japanese kernel I don't know how it would impact if it would be installed well maybe the way it would impact in the store is that if we had to wait until 268 then we cannot I'm not going to push for that for edge because we don't have the support for that kind of security in Japan yet anyway so this is what he asked about if you're going to do something about x and a half yes there is at least one set of people who would love to have the if you don't have the support in stable at that time it just makes sense to have I don't want to push it for edge because I want to freeze things so I don't think I can do it between now and December the packages that I'm talking about that I have to update they won't be frozen until late late in the fall so wait in the fall what are you talking about at least in December so it's a Chinese by the way generic it's a multi part of base it will be frozen in July or August and an old package will be frozen in mid of October so my the way I see it right now I want to get the basic I want it working with the end of August I wouldn't have the multi level security working until then they're being installed so when will you have that I think I will have that working by July or August I don't basically my understanding currently these are really optional part that I have a very from the outside point of view are optional or extra so we could extend by late and if we work it's good and if not it doesn't work in the back and also if you say okay we need to update SAE news policy because it's broken I might because I shouldn't do it in the policy later because they are not in the co-part my question was it's going to be out in December there's no point in discussing it's not going to be after December with August 2018 I hope it will be out in fall December when is it 6 o'clock for H it needs to be not all decided but installed bit of true light because if you want to make a DEI a release candidate the I release for H in the beginning of August there's no point in doing it later now that's right although given the schedule I think we can do one or maybe two six sixteen updates after that because rebuilds of the I from the curve are much easier now we don't have to wait for the individual builds of the curve we just have to coordinate with supporters to do the unit builds quickly and I think we do have the operation is that good for all current updates for all current updates in one current series like two six sixteen last can I explain a little bit more about I think even though the official current will be two six sixteen for H there will be two six sixteen as well two six sixteen will be out before we release H in our timeline my feeling is I'm creating here KML if two six sixteen is out the extra and the optional packages that support the feelings in H will support two six eighteen so which is why if there is if the kernel team after we release H officially releases two six eighteen then the security users don't have to go off on their own so that's what we're trying to do okay so I was just notified now by the heads of this event that I should just come to the end soon I always think we have done lots of important stuff so it was mostly about the kernel but obviously the other thing was really interesting to most of the users otherwise I would just say thank you to you all okay we will need to talk about a new kernel because it does have implications for you certainly I have to talk I have to talk to the kernel we just as late last night when we talked the kernel team is not in agreement to do it you would also need to see what are your main packages going to report to two six sixteen or two six eighteen otherwise it's all the main packages so the usual and not what I'm going to point to but one question for us about that wouldn't it be possible to make if you don't have a new kernel we need to build it in the store also if you use the kernel to do installation so if you do that could we say please don't install this is make a package for another make a package now you have to basically if you want if you want something else that wants the current make a package for instance you have to do a medium or long priority install so you have to do an extra make a package for another if you leave this in the store could we change it so that we could install another make a package or is it to much work depends on the name of the make a package what else needs something I would say it's not very nice something I was thinking about was the way we do like 2.6 and 2.4 today that we when we build the I if we do this in the future well the project is that the kernel major is not going to change and we currently use the kernel major to decide on the make a package I suspect that if we do this the kernel major would change the ABI would change oh I see what you're saying kernel 2.6 but because I have to make a meet a package I must say it's ok this is now the kernel kernel edge it's a package it's a place where the system a package after it's a model release so the users will get updates with the next this update after edge but in the old installer the default installer but you created an F3 you wanted to succeed but they had separate majors so that was easy so we based that decision on majors we would do a consistent installation if you build the 2.4 we would install 2.4 if you build 2.6 you would install 2.6 to extend that to a miner it's going to bring things horribly especially in stable because there are just too many projects of miners are you looking are you ready could that be a digit dot digit or can we have a better package or install for edge that's yes that's a good one you should have said kernel 2.6 edge no no you should have a linux image edge whatever I don't mind about names if something is useful we are of course willing to think about it but it's not something I can just do it's easy it's fine of course you have to discuss first you have to discuss first ok that's a nice thing I should now finish this set because I don't see more and more signs as I get nervous so we finish thank you all for coming what's good to start with yeah that's a good one