 Sorry, guys, for being delayed, but we are just a bit experimental. More experimental is in my least talk, for the least talk I was just on time. This one is a bit delayed, but experimental does it quite too often. I hope we have an understanding of that. So, yeah, what about experimental distributions? We have in WN 4 or 5 major distributions stable, testing unstable. We all know them, we all use them more often. Aber die User all complain about them, because that's the usual ones. We have another distribution that's called experimental. In experimental you can just upload packages that are not meant to go to testing. The great advantage of experimental is, if you upload anything, you definitely won't break our testing migration. So, speaking from the least end point, I'm very, very happy that experimental exists. Of course, in some cases, if you have just a leaf code that nobody else uses, you can of course use unstable as a station area. But if you have a nice library, for example, if you are a KDE-Maintainer, you better not upload to versions 3.2 Unstable, then we declare that need please. With some or just that. But he's not KDE-Maintainer anymore, so it's not so bad. So, yeah, but for experimental, our usual build demons, such as build packages in old stable security, in stable security, in stable post updates, of course, in testable post updates, and in unstable. Well, this is of course the optimal thing, because if you want to stage something experimental, you perhaps want to see how the build demons behave on stage in that. So, we started through this building packages in experimental in, I think, August last year. We actually started only with, I think, about four architectures, but now we are up to eleven architectures. I don't think that's right. I'm in the sixth floor, I'm not so sure of myself. So, it's an outside build, but all the other architectures, such as build usual things. So, it's, well, so, it's one thing when experimenting is actually much easier to build in unstable, because if the build demons are somehow broken, nobody really complains. Of course, they start complaining if it's very broken for most of the weeks, but nobody notices if it's broken for one week. If the table would happen with unstable, we all know what happened, what happened at the end of the world. So, it's actually also, not only for the package, but also for me, also a very good chance to learn to know how the tool changes work, how the build demons work, how the wanna build works, all the small stuff. And now I'm using the knowledge that I've learned for auto-building volatile, for auto-building test and security. So, I think, well, this has been helped us a lot. Also, some, well, just for how the experimental auto-building usually works, we have a wanna build located in the University of Darmstadt. Just the same wanna build software actually is used on unit for the usual Debian packages. After each mirror push, um, ja, wir haben also ein paar secondary mirror on the University of Darmstadt, right now. Und es tun, es ist mir das Update, das heißt, mir das Push, ich push das so an, a build, to update its package database, denn es makes a usual comparison. Usual comparison means it's an squinted, which takes a packages file, it's a first file, looks, which packages are there, which packages need to be built. Gau in das wanna build database, es bildet immer so aus, wanna build, und say, okay, ich wasche das Bild, ja, wanna build, und das buildt immer so aus, building it. So, tu was sie additional, ich schuss, connecte ich das Auto-building experimental. Vor Anstabil, the other thing is quite easy. You can build a package, you can upload it, it's fair. It can depend on anything, set's also in anstabil. For experimental, it wouldn't be too good, because, for example, if there's a new libc version in experimental, which has had, sometimes, some of the last three sites, and the auto-builders will build against that version in an experimental, which has had an ssh-Bump. Well, anybody who wants to install the package is sent for us to install the experimental libc, which is perhaps not what we really want. So, vorwärts zu sehen, our build, at the very beginning, our build, it just started to only build the experimental package in anstabil, so it means, if they have any build dependencies, they need to be, we need to be, we need to be possible so anstabil, so as to fulfill them, otherwise it will fail us. Now they have been some testing of changes, so it means, has done some code changes, high-out, by the way, has done some code changes to build them. So, as the build team noticed, oh, now I need something from experimental to build against, so is it all from experimental, but all of what is needed from the build dependencies. This works now quite fine, but of course, there's a second problem, connected with that. For example, if an usual build team gets a build dependency on a new app, it just installs a new CC version, and it's there. It's never uninstall, because it's build essential. If someone happens experimental, well, yeah, won't go do it, but have a lot of fun with it. For that reason, you need a usual version of the change route with just the version packages. It doesn't work as expected. So, for example, one thing that I use on one of my build demons is I have two change routes and the first things I'm back after finishing the build. This advantage is, of course, you need more space. In some architectures, space is really a limiting factor. And by the way, first thing is not to plot to pass this program and some architectures. So, other ideas that is currently being applied is to use LVM Snapshots. So, for example, if you just make a snapshot, build it or roll it back to the last snapshot, I think that's just things that are still working on. Outputting experimental is definitely not something where I would say, it all works. It's all done. But it's more something where we are working it, where we can't line up new things. I think we have been, in my opinion, outputting experimental works better than I expected. But it doesn't mean that everything works. But on the other hand, I think it's really nice to have some space in the area. In Debian to try out build new patches to try out wanna build patches without it's a great danger that you first try like just unstable. Okay, so, I think that was now enough for some introduction to the topic. And I think it should be more, it's a bit of a talk. So, I think it should more like a discussion topic. Yeah? I think on the other side just switch the tips. Ah, okay. Yeah. Ja, of course, sehr, was also exist is, of course, a bit as a database of the build logs, the usual build logs, said, okay, I was asked from Martin to go on the other page. So all of a sudden he asked me to repeat it. So, I did it. Yeah, so. So, yeah. So, we take more or less the same software that's used and build it Debianhawk also so was a buildbox set up with use. You see, still some, a lot of fails. For example, we go in in the first package flasche ja, in this ttg-bind one for EA64 so take the direct link to it. So, well, I also know why it failed. I don't even need to look at it because I definitely need something newer So, jetzt ist das was ich erwähnt habe, hier T-Tage Base, es ist einfach nicht so, wie es sollte sein, weil es auch vom Experiment etwas anstabilisiert ist. Das ist eigentlich der beste Kommenbild-Felcher, das ist das Bild, das ich habe, das Wouter, das Heldbewerb, das Heldbewerb, das Heldbewerb, das Heldbewerb, das Heldbewerb, das ist kein Problem, das es solved, weil es dann in diesem Case die T-Tage Base wirklich betägt, vom Experiment. Also, auch das 2. Sieg, das hält, weil hier ist das Bild, das ich habe, das Heldbewerb, das Heldbewerb, das Heldbewerb, das Heldbewerb, das Heldbewerb, das Heldbewerb, das Heldbewerb, so knifecupt hat es zu�� 최고 für die T-Tage Base, also die T-Tage Base, also die T-Tage Base, diese ist die neue Phase für die AT Detail, die es damals Also ich habe jetzt noch eine Pizza-Odion, also ich habe selbst das KS-Bus-Wonder-Test und Versionen, die ich testet habe, in einem von meinen Bild-Demons, aber es ist nicht so ein Sieg mit den Kurren-Versionen. Aber ansonsten ist mein Kurren-Time-Schedule, wenn es noch bei der Release auf Sarge ist, dann wird es nicht mit den KS-Mails aufgelassen. Also mein Bild-Demons ist aber die Kurren-Solution, die wäre besser in der Note auf andere Karte. Ich hoffe, dass man das auch mit einem Bildschirm hat, und dann, z.B. das Package, das hat funktioniert. So, ich denke, jetzt ist es Zeit für eine Frage und eine Frage für dich. Ja, wer möchte mehr, um zu beginnen? Das ist die Kinogestellung. Also, any of you? Ja. Do you have approximately what percentage of experimental build successfully versus not building successfully? I guess that it depends a bit on what is curving experimental. Okay, first I would like to ask you a question, of course. Please put the sign up right up here. Ja, weil es ist liegengeword, but never mind. Ja, the question is how much percentage of builds an experimental are successful and which is not. So, the answer is it depends a bit on what is curving experimental. And what kind of failures. If there is one failure like this case, we will just say it can't build for the reasons, that we wouldn't experiment to build in, what we don't do for the pencil reasons. I guess that there is something between 10% and 70% depending on which time. For example, it's a time with known to aid was an experimental. Well, Gnome has a lot of packages. They are all dependent on some Gnome level that was on an experimental. Ja, lot of things fail. It's a fail so much that it even consigned to go to any of my CH routes. Just install the reliable package of experimental and then start to rebuild. If it takes a structural failure aside, I guess that about 5-10% of the packages fail for the usual reasons that are forgotten dependencies, package only builds on the maintainer system or the kind of stuff that you see also in Anstattel. And if I see such bugs, I file usually a C-Box tagged with experimental so that the maintainer is encouraged to fix it. Because usually maintainers don't see it by themselves except you tell them. All of the maintainers tell you, oh your package is still missing, my package is still missing, why didn't you build it? I tried to but you didn't do it right. Okay, next question please. Do you know approximately how much is experimental actually used? I mean, how many packages are in experimental? Oh, so the question was, how much is experimental really used? Ja, really, well I can't say, set number is currently now in. I don't even know if you can hear the question. Also a question that's a bit difficult to answer. My feeling also from what I see from the build logs is that the usage of experimental has been decreased very much in the last five weeks for obvious reasons. It was used very, very much before the release. There was app.06 installed in experimental, which a lot of people installed in their systems because it's a secure app. Even now a lot of people have secure apps from Anstattel, which has installed in their Sage-Systems. Also, it was used very, very much for the GNUCC Transition, so this is working right now, because there have been prior uploads of GNUCC 4.0. But that package was too large for us to build some architecture, so some just failed for lack of this space. Ja, I hope it's the inverse. Now you're laughing about it. Ja, so I think it's really of some use and it's of very large use to really see, okay, I have some package, especially if you have a large easy package, like a normal library, it depends on how stable you can upload it and see, does it work on everything, whether I texture it or not, before you break some transition to testing. So, okay. Wauldrich has currently 202 packages in the database, so I think that's a fair share. Okay, next one, please. Just in addition to your... Wenn du an Edition bist, komm hier an und warte auf die Fiete. Just in addition to what Andi just said, that we are running out of disk space, that problem should be solved very soon, because we got promise that we will get some more disk space the next week, so... Ja. We'll be just noticed, that Bob in his basement just some laying around was. So, we all know he has a great basement with a lot of computers, every chance I expect too hard is from there. Ja. But something also, well, you also, it taught me a lot of things about running build demons. For example, for getting to the testing security thing is, with biretorities of storage, getting to the testing security, the change suits up and the auto building experimental helped us a lot in it, because it just tells you what the issues are and if somebody comes to you and says well, this thing doesn't work now with running build, they say oh yeah, I know that. I have the same issue before and the solution is that one. Ja, so, a rather great way to learn things and you can't do as much harm as you do if you just do something wrong with a building unstable. So, next one. I don't buy it. The heading there said something about building backports.org, could you say something more about that? Ja, of course. Backports.org ist supposed to be out of the picture. From my point of view it works, but Backports.org doesn't use the usual Debin-Archive-Kit, but some self written software and that one has called Debin-Archive-Kit except EC86. Also, the maintainer of Backports.org is currently taking some vacation. Of course is of course very okay for him because shortly after the release Backports.org is not as important as it was this shortly before the release. So, that's okay, but as a current player is Jörg Jaspert Ganev who is also an FTP-Team und I will install Debin-Archive-Kit for Backports.org und after that is successfully done we will auto-build Backports.org for, well, it depends something like 11 or 12 architectures. Definitely we will auto-build for PowerPC and for AMD64 and for the architectures we have an up-to-space in the University of Darmstadt and well, in der Bild-Dimension. One thing I really should say is I'm not doing all this myself. I'm just doing the Savannah-Bild and doing some of the Bild-Dimension another Bild-Dimension for M6AK is a bit louder. Another Bild-Dimension by Kenji Emuto is a bit farm. So, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so weil er zeigt, dass wir das Lande bilden können. So, Elmore acceptet die Plätze, weil wir prüfen müssen, dass es funktioniert. Weil ich könnte sagen, okay, Emma fragt, du kannst die Probleme nicht erinnern, wenn etwas losgeht. Ich sage, ich weiß nicht, ob etwas losgeht, aber ich habe mich jetzt seit neun Monaten in die Issue. Und ich sage, okay, wir können das erinnern. Also, es ist eine Plätze, in der wir wirklich die riesigen Dinge sehen. Ja. Okay, did I miss anything from the answers to the question or the answer all? Ich habe es vergessen zu sagen, sorry. It's hard to say the question into the answer. Especially you want to, about backwards work. But perhaps I should also say, what else do we all build, the same build that we have to work. On the one hand we want to build, but I see, there are still 50 minutes we have to fill. So, originally we started as experimental. So, some were limited set of, on some packages build and some of the build demons. Some build demons, packages build, York Gasport has an experimental deck, they have a kit installation for new maintainers. And some build demons. Especially it's not on the IA64, even because that one is located in the US. I don't want to group to export anything from there. Then Volatile is built on the build demon network. And we are about to set up some new archive for tested security stuff. And they will probably also build on the same build demon network, but still work going on, and we will publish it as soon as it's there. Okay, you? I've heard a rumor, the experimental spark build demon is much faster than the regular spark build demon. Well, that's what happens. Well, actually we offer our build demon to Elmo. And he promised me to get back to my main, at least at the end of next week. And to really do it this time. But the reason why it's fast, is because it's just easy. The University of Dungeon had a lot of spark machines, and we said we're just standing there, powered off. And of course we took the fast spark build demon. We should recheck something lower, something not as fast as possible. Actually, I really divide my build demons into two sets. Some fast and some slow build demons. In the spark one is a fast one for me. It's a Power Passie, or the EA64 one. Slow ones are, for example, the most build demon I'm running. Especially slow is the SN90, but that's because it's running in the Hercules. Yeah, next question? Obviously nice, because you don't have a breaking stable. Is there some easier way to... I mean, you still have to wait until the package is picked up, and it goes through reinstall and stuff. Is there some way to get a package all built more or less instantly? You don't want to test something, so you're actively short of mail and admin and get in and store all these packages? Ja, all these don't exist anyway, was I'd say. Currently, es wird definitiv nice, zu erstressen, auch an setup Auto-Bilders. Wir could just say, I upload a source file in a specific, which architecture I want to see, as a build block. Ja, that would be nice. But, I see some danger in the setup launch also, vor sehr simplen Sießen. Ich meine, ich meine, nur ein paar Leute, die einfach starten, um die Send-Source-Package, also die Uploader anstabilen. Es ist eher wohl vor allem der Architekt, sondern eine Turn-in-of-Bild. Es sind sehr einfach Uploaded. Ja, es wird wirklich nice, so wie eben du es haben gesagt. Aber, ja, ich kenne du, es sind Pups, da ist es, haben wir auch mal auf die List. Aber, ich hasse nicht zu warnen, du. Es ist ja eine Menge Positionen, die du auf die List hast, z.B. da bringe. Don't have a fair Scheduler. Ja, es hat, okay, ich habe jetzt viele Fragen. Was ist die Interaction zwischen Experimental- und Backdeckingsystem? Okay, ich will describe, perhaps, first, how I will look like in the future, because in the future we are about to have version-Dreckings-Support. Version-Dreckings-Support und Backdeckingssystem. But, I don't expect it to happen fast in the next two months. Because the people, the person, who should deploy has just very other real-life priorities. Yes, it happens sometimes, except if one knows how long it will take, it's of course fairly okay. So, that's okay. And if we have that, a bug will just be marked, okay, this bug is valid for this version of the packet, to that version of the packet, and you can have multiple markers, so multiple start and days, like, oh, it's hard to be, the third version is disabled, to that version is disabled, and that version is not an Experimental, is that an Experimental? So, then, there will be a very good Interaction. The current status is that I upload back a bug tagged with Experimental, and what Katie does, or all that Katie does, is, if you upload something, if you upload something to unstable, bugs are closed, except if it's an NMU, in that case, bugs are marked fixed. If you upload a packet with Experimental, bugs are marked fixed in Experimental, if it's a maintainer upload, and a marked fix if it's a non-maintainer upload. Which, of course, it's a bug that holds with a tagged Experimental and fixed in Experimental no other sweet thing. That happens, but usually the maintainer notices and glows them. If I notice that some of my bugs were set to that state, that was definitely not valid in anstable, so dass ich das glowset. But sometimes I just don't notice it, especially if it happens at a day, where I just send something like 50 Comments to a bug-tagging system, to just some other SD-Bugs to upload. So, yeah, but it mostly works better. Okay, next one. Okay, if there's no next question, I would just say we could finish a bit early, so we have some more time to get used to the next talk. Okay, so, I just see now without 10 minutes left for an extended break. Okay? So, thank you very much for coming and for asking questions. And I hope, yeah, we all learned something on that and we get even better with Experimental, with building Experimental in the future. Thanks a lot.