 I gael y gallu'n gweithio, gallu'n Gweithio Mark Góvern, ac yn ei ddweud yn gweithio Lenny, fel y brotech yn rhoi. Ond ydych chi'n gweithio. I gael y gweithio, wrth gwrs, yn ddweud yn ddweud i'n gweithio'n gweithio. Roeddwn i'n gweithio'n gweithio ar y cyfnodd rwyf ar gyfer hyn ymgyrch yn ffin. Felly ymgyrch yn gweithio ar hynny. Ond rwy'n gweithio'n gweithio ar y brotech hyd yn Gweithio Llenny. ac ydych chi'n gwybodaeth ymlaen i gweithio i ddweud i gael gydag, ac yn gweithio'r gweithio'r gwaith, sy'n gweithio'r gwaith neu'r gwaith. Dyma, mae'n cwylwch ar ymlaen, mae'n gweithio ar y gwaith yn y gwaith. Mae'n gweithwyr yma, mae'n gweithwyr ac mae'n gweithwyr ymlaen. Gweithwyr yma, mae'n gweithwyr yma. Mae'r ddŵr ffordd. Mae'r ddŵr ffordd ddim yn gallu bod ffwrdd ei wneud o'r ddŵr, ac y byddai'n gwneud o'r wlad o'r byddau o'r tymysg, ac mae'r ddŵr wedi'n gwneud o'r dŵr. Ond oherwydd, mae'n ddŵr ffordd ymlaen o'r ddŵr. Oni'n lefwyr ar y dŵr, a'u sydd ydydd yn dda. Mae'r ddŵr yn cael ei phawr o'r ddŵr, oherwydd o'r ddŵr o'r dŵr, Penedigwch, sy'n gweithio i gyfnodd yn gweithio ac mae'r cyfnodd yn gallu llawer o'r tyfnodd yma. Llanfa, mae'n meddwl yn fawr o'r gweithio. Felly mae'n cefnodd Aber a Vallon. Ond yw ynogi'r gweithio ar gyfer o bobl yw'r gweithio, ond mae'n fawr o'n gweithio i gaelio'r cyfnodd. Ac rwy'n gallu ei wneud dwi'n gwneud eich bod yn du yn gwneud y magwyddiant fel y dylai'r nesaf o'r ystafell yma o'n ddod yn dweud y mangylcheddau a'r ystafell wedi'u gwneud. Yn gyfnod i chi, oherwydd. A ffynol, rym ni'n golygu'n ddod o'r ystafell ar y cyfnod yw'r ystafell. Rwy'n gweithio gyrfa gydigol, prosesau, gyda'r hynny, mae'n gweithio'n gweithio'n chymlach. Rwy'n dechrau, mae gwybodaeth fydd ymwyl gan gyfnod y system unibild, ac mae'r straff iawn i ddweud yn gyfoglu Fbackham Cymru fel mae gennym ei sylwg o dim arnyn nhw. A gwasanaeth perio i gael fyfyrdu cysylltu, eich llwyaf chi'n gwybodaeth. Yn ddweud, mae'n ddigfael. Maeeth yar o reswm. Mae'r rath gyddoedd hynny angen. Wrth gwrs, mae'n bwynt. Ond rhai, dwi'n cael eu cyfrosech. Nid yw'r ffordd sy'n cael ei ddim yn cael ei ddisloedd ffordd sy'n cael eu diolch yn ei wneud. Fodd yna, rhaid i'w gweld i'r gwahod o'r ddweud ym mhwylo'i gweld o'r ffordd sy'n cael eu ddisloedd. Mae'n ddiweddol i'ch ddweud o'r ffordd sy'n ddweud o'r ddweud i'ch ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud. ydy'r hyn yn rhaid i gweithio amser y graff ac i'w ddweud yn ffriis. Fe ifrwch, mae'r byw wnaeth ddweud yn y gwaith deapo, ond mae'r draffan â wnaeth ar gyfer, gallwn eisiau yn ei ddaeth yn ei ddweud ar'i'u ddweud. Llyri wrth gilydd, ni'n i'n cofnod i'r gwybod, dweud y bydd heddiw rhai yn ymwheel o'ch beth rydych yn gyf ôl i'n cymdeithasol, ac iddyn ni'n cael gorfod o Gymru yn digwydd y bydd y bydd. Ac gwnaeth ei shellwch arall y tannwg deisiniad a'r tannwg reichu'r tannwg a'r tannwg gwybodaeth. Rwyf yn rhan o'r holl yn gyd. Mae'n ei gweinydd yn y bach o gwybod mewn gweinloedd, am amdano y cyflinkwadau cyflinkwadol yn cael ei gweinydd, i wych yn y mae'r cyflinkwadau achos cyflinkwadil acnych llwylo cyfo'r arbedig. If dyfod eich cyflinkwadau a'r cyflinkwadalalol that you want to make sure gets into, that the fix gets into Lenny, then you should talk to them. I think it's testing-security at listslotsdewing.org. So normally we can offer advice on what to do and help get this fix into Lenny for you. So, unblocks. This is what people are mostly asked for on our list anyway. I should point out that this image was one of the first ones on Google image search when I searched for unblock. Llywodraeth hyn oedd ychydig wedi gwahodd yn ymdilloddiad fan hyn. Felly daeth онwyddog am bwysig ychydig i ddisgrifennu'n ddysgu'n blynyddiad ar eich credlwn. A ti ti rodi wneud gwneud ei wneud i ddysgu'r sefydliadau'n ben ffathfforddau arall, yn y Cymru a'r byddau. Rwy'n credu bod yna. Felly anodwch i'r pwysig, fe yw siarom o ddyn nhw sy'n gweithio. Fy hwnna gydag o holl, Felly mae'n gweld o'r listfyn i g dorllun i leisdog, dyfodol i'rhell yn cyllidol i allan. Rwy'n meddwl o'r posibl a'r posibl, mae'n meddwl, ond yn mynd i efallach yn ddweud, rgynno i ddiweddysg peth, ychydig yn ddych chi'n meddylch. Roedd mae'r trofyn yn cael trofyn a'r troll yn geisio. Roedd roi'r trofyn yn ei hadegol, a oedd nhw yn gweithio'r troll yn gweithio. ond we find pre-approved fixes, so stuff we've already said okay too. Now to get that, you basically, if you talk to us on debian-release at list.debian.org, then you can get, then we, it's a good way to get any particular fixes you have in mind. Generally the most important thing is just to talk to us and get information about what you'd like in there. So how to ask for changes? Generally you may all debian-release at list.debian.org. I will be repeating this quite a few times because this is the most important thing I can stress throughout this talk, is that you talk to us and we are able to work out what's going on. Having random patch sets arrive and we have no idea what's going on and then people ask us to unblock them, doesn't really work too well. So don't make, first one is don't make unrelated changes. You should only make changes for the fixes that you want in, so don't rebuild the package, change to CDBS from debhelper, introduce a new patch system and yes I have all seen these in the past week. And explain why you need it. So we want to see bug numbers, deb diffs, change log extracts, just saying please unblock X is normally not good enough because we have no idea why or what it fixes and it creates a lot more work for us at a time when we're all really quite busy anyway. Normally mentioning the package name is helpful as well. I'm not going to mention any names but normally telling us what you'd like unblocked is quite useful. And again if in doubt just mail us and ask us. We also have an RC channel which I'll get to in a bit but the preferred method of contact is via the mailing list. So what you don't get unblocks for. Things like the top one where you have 12,000 insertions and 7,000 deletions across almost 100 files. This generally doesn't get an unblock unless there's a very good reason. The issue with unblocking things is we have to review the change set between each of those and if we have that much to do generally we'll just say no because we don't have time to go through all of that unless there's a very good reason and it's been very explicit. Bumping debing compact can cause interesting effects even if you go from very low number to the latest number. This does change the build system and how things work. It's generally frowned upon to change that unless it's been extensively tested. And again you should mention this in your mail source if you do have to bump it for a good reason then just tell us and we'll see what we can do with that. NMUs seem to be overwritten. Try not to redact the change logs and ignore previous NMUs Try not to acknowledge them and use the latest versions. For example changing the build system again is another example of something that shouldn't be done. New binary packages introduced new upstream versions. Generally try and keep your change sets and the delta between the two fairly small and then we should be able to include it in the next release. So if we have what we call stuck packages items which normally we prefer everything to go via unstable sometimes this obviously isn't possible which is why we have testing proposed updates. So sometimes you can't put your package through unstable. Examples of this would be if there's unstable changes already in unstable like you've got a new three new upstream versions sat there compared to the version in Lenny and it's just been uploaded and there's a problem or there's a big library transition and there's an incompatible change rate. Now to get round this you can update via testing proposed updates as I mentioned which will put an item straight into testing so it won't go via unstable with normal routes. It will get auto built but it's very important that you contact us especially as if we try and if we see a package come into testing proposed updates and we're not sure what architectures it's built for or get a bit confused as the way testing proposed updates works currently is that it will just put the package straight into testing without checking that all the architectures are there and then you end up with quite a broken package. So it is very important that you talk to us and tell us what's going on and check that we can put things straight into testing for you that way. So removals this one's caused quite a bit of discussion recently on various lists. The most important thing to remember is RC bugs equals removals. Just make sure that your RC bugs are fixed or you're mailing the bug regularly and that we can see that some progress going on. Optional and extra if it's RC buggy for greater than a week it's subject to removal. Again for the others if it's great if it's RC buggy for more than 20 days and there's no activity if it's a leaf package etc these are all subject to removal. That said there is no guarantee that any package will necessarily ship with any. If it's buggy or if there's other problems with it then it may get removed. It's important to note that we don't take removals lightly there's always various reasons for it and there's discussions on the lists that I'm sure everyone will be able to find about why packages were removed. A new policy which I'm not sure is officially up but it's all been brought up is that if it's been orphaned for more than two releases this is similar to the new FTP master policy of dropping an archive out of the architecture if it's not been there for a couple of releases. A new one is we had a look at the huge amount of orphan packages that just haven't seen any updates or any uploads. It doesn't seem anyone cares about them because no one's maintaining them. So if it's been missing for more than or equal to two releases then it's subject to get dropped out of the archive altogether. Now if a package has been removed recently and you have a particular update so if it's been RC buggy or failed spill from source and you have a fixed package that does this if you then upload it to unstable it will be blocked and it won't be going into testing. That's normal. If it's fairly recently so we can see it's only recently gone so there won't be too much regression. If you mail us an ask us again we can always get it pushed back in. Once again the important thing is mail an ask us. Now I keep talking about these RC bugs because it is one of the most important things we have for Lenny. It's one of the main criteria that we shouldn't have any RC bugs when we ship. And this is the sort of current status. We do have as of last night 378 RC bugs still open which is quite a few really. If we have a look at the quick graph it's that. Now this has led some people to work out when we'll release and try and work out various states based on that. And the current thinking is 2906 which is a little bit delayed from what we'd have preferred thanks to that man. Mr Bastion anyway. So the realistic question is when do we release? There's a couple of different methods of trying to work out when we'll release. There's our original timeline which says we'll release in three weeks. Now with about 380 RC bugs that's approximately 130 a week or so. So 20 odd a day. So who feels like fixing some RC bugs? To meet a three week deadline we'd have to fix about 20 RC bugs a day and not get the rest of them done. Which is quite a sobering prospect I certainly didn't realise that we'd be quite so up against the wall in terms of time. Unfortunately the various announcements have said we're due to release internally. They've said we're due to release in September. So that adds another couple of weeks. Five weeks. Still not too long. The official release announcements have said we're due to release in the second half of 2008. So we can still meet that by quite some way. But the most important thing is to get this RC bug count down. So to get them fixed and not introduce any more hopefully. So a bit of contact information. There's firstly our website which is release.debin.org. This contains various bits and pieces of information, our policies, a little bit about the architectures which I will mention. And what release goals we have, what we consider RC buggy, et cetera. There's mailing list which should be a primary contact point for the release team. It will allow you to ask for unblocks. It will allow you to ask questions. It's the preferred method of communicating with us. And we also have the hashdebin release channel on RC.debin.org where there's normally a few of us around. I should, though, point out for unblock requests should definitely go to the list. There's various information in the topic there that might redirect you to various other places. So please read that. But if you just want to inform or chat or just to work out what we're talking about, then that's a useful one. And unfortunately, with 15 minutes in, I seem to have run out of slides. I did say this was a rather rushed up talk. So I suppose I should open the floor to any questions or any particular concerns anyone has. If not, I will be dragging up one of the other release assistants to give some more talk. Ah, we do have a question. Thank you. I have a question because I noticed that I was using one package on a server and was not aware, on edge, that it was removed for Lenny because of RC bug. So I took over the package and it's fine now and it is fine for me personally. But it shouldn't be a release goal to take at least those packages which are running on edge to keep if they are now in a good shape but was not before the freeze. So the question is should we make sure that we don't lose any packages between releases? To enable smooth upgrades for releases? That's an interesting one. If we carry on along that set, then we can never get rid of old packages. For example, the GTK transition from GTK 1 to 2. Some packages just don't seem to be interested in performing that. For example, XMMS is going to be stuck with GTK 1 and we can't keep that released. Of course, we do like to try and ensure that functionality exists between releases or equivalent packages are there. You can run on your own system if you are in Lenny. I think it's RC Alert, which will tell you if items are RC buggy. I learned this functionality just yesterday. It is great that we have the script. That could be documented better. If there is a package that's really important to you? The situation is different between the GTK 1 issue and the XMMS issue. It is fixed not yet. These packages are broken anyway. If we now realize that we could get a package in good shape which was an edge and just was lost in space and nobody cared about it but we have now a good package. Would it make sense to put it in the release or not? It may be possible. It depends on why it was removed. If those are fixed and it was fairly recently removed. If it now seems that there is quite a lot of interest in the package then it could be reintroduced. It is not a lot who is interested. It does have to be taken on quite a case by case basis. Thank you. There has been an idea about what Andreas just said. It would be to add a cron script in Dev Scripts. Every developer would have it running on his machine to monitor the new RC bugs and the newly opened packages amongst the package that are locally installed. That would allow everybody interested in the development to have an overview of what is currently going to be broken amongst the packages that he uses. Currently it's true that it's really hard to have a global view of that. Currently it's quite controversial because people say that we shouldn't add a cron job to Dev Scripts. I'm not sure what the majority of people will think probably it will just be a matter of the one who can do it, and it will be disabled by default or enabled by default, but you could disable it. Yes, I certainly remember that. We certainly used to have a list of RC bugs go to DDA, I think it was, until there was complaints that people were getting too much information about RC bugs, which I'm not quite sure how it's so important if you can't get too much information about it. One of the main things I would like to push is that release critical bugs mean that your package will not ship with the next stable release. If you have an RC bug it is important that it gets fixed, and it gets fixed promptly. These are bugs that are so severe that the package will not ship. It just won't ship, we won't support it for the users, so it is something that does need fixing. To try and cement that in minds of everyone would be very, very useful for us. Again, this is now in, I believe it now goes to Debbie and Develle. Does it? Sometimes it goes to Debbie and Develle. I'm actually, that's Steiner Gundersen who's done most of the hard work on the RC bug alert, the debug side of that. I'd have to check again where exactly it's going. The reason why we didn't send it to DDA is because it, I think it was the WMPP alert, was all stuck together and people just drowned in it. They didn't actually ever look at the RC bugs that were there. They just sort of said, yeah okay, it's the same email again this week, delete or mark is read and no one really paid much attention to it. Rather than flooding DDA with information that no one looked at it was decided to split it would be better. One thing I would like to try and get done is once these RC bugs get low and get to zero so we can release is then try and keep it at a very low level. So when RC bugs do occur it should be quite an unusual event and they do get fixed quickly trying to keep this count low so we don't go up to a few above 500 or 600 or so because then when you get to that sort of scale it just gets lost in the noise that there's so many RC bugs and people don't know where to start. So trying to keep these at a manageable level and make sure the RC bugs really get fixed as soon as possible could be useful. Yeah I could go back to the dev scripts RC alert. One of the other things that you can easily do is just look at the overview of your own packages at qa.wm. I mean I assume almost everybody. Does anybody here not know about that? No okay I didn't think so. So I mean far more useful than the cron job because you can look at it when you have the time and I don't know I have dev scripts installed on about 20 machines each of which has mail. I think that's the problem that we're trying to solve here is that most developers only care about their own packages and we're trying to get developers to care about at least the package that they use but they're not maintaining so that would be the point of having such a cron job because on your DDP page you only see... Yeah sure it's easy to write yourself but if you provide it a dance point was everybody could write it Yeah okay I know it's wine line but if you have that script you can then improve it and get more useful information in it like the list of packages that were removed from testing and that are still unstable and that you use. So well and if you provide it then people will run it if you ask people to write it and to enable it themselves some people will do that but most people won't do just a... perhaps not. Okay yes as Lukas said you have to have this script and you have to advertise this script because until yesterday I wasn't aware that there was a script and I just found four packages I could easily take over which do not make much work so I just use it and they are often and if we would add to the monthly posting which work should be done I don't know how they are called just run this script or if we enable this by cron job or make some more noise about this we could get much more input by developers who just not know what happens in this case. Okay so there's various useful scripts we've got I mean how many people read the release updates that we send to Devin developer now hopefully once a month it's quite few and how many people read right to the end of those a few less. We always try and include now a little snippet at the end with hints for developers that mention various hints and tips so we've got various scripts and items like that I'm not sure if RC Alert has the RC Bug Alert has been included but certainly others have possibly not putting them right at the end is the best but that's normally where we release normally where we announce the code names for the next releases as well so we do try and keep a little bit of motivation for everyone to read them. Is that working? Okay. Some another topic I just mentioned a comment by Zobel on IOC he mentioned that maybe not all RC buggy packages will be removed such as build-ups essential or such stuff anyway that was a side comment another topic I wanted to raise I'm not a specialist of this but I have not heard about it is the kernel and DI what are their place in that release process? I'm quite worried about these topics currently DI has been released quite two months ago is not yet released and not released with the 2.6.26 kernel so how will that fit in a three weeks process? Yes, again this has also been as I'm sure if people read the release list and the DI list and demo boot it's quite a contentious issue working out we have three different camps which are involved quite heavily in a release we have the release team which is obviously managing the general infrastructure of the release and what packages go in we have the DI team who are managing the installer and we have the kernel team who are trying to stabilise on a particular kernel now within those three teams we can easily have 6 or 7 different opinions as to what we should do especially with the latest what version of kernel we should have and how that manages with the DI team at the moment certainly the release team we try and work closely with the DI team as closely as we can The main problem I see is that the DI team is currently shrinking a lot I think that some of you have seen Joey's announcement on his blog Franz Popp also mentioned some stuff that it will not work that much for the Dylan Lenny release so I would like to raise something here DI is not that ready for release and we probably need a lot more manpower and testing, testing, testing anything for DI because my current feeling that's maybe personal is the current DI is not as good as the H installer was it's probably more buggy so we are likely to release with a more buggy DI if nothing more happens and Otavio is doing a great job Jeremy Bob your also is doing a great job with DI but mostly alone which did not happen in the past in the past there were more people involved in DI So one particular question would possibly be working out why at this stage three weeks before the planned release date it is buggy I don't know was it a surprise that we should be releasing actually this time of people being assuming that the deadline we set wasn't real deadline or what? I don't think there is any surprise here this is more or less something we could suspect already one year ago that the manpower in DI was already starting to decrease and well this is something you can do that much if the manpower is decreasing you cannot bring easily more people around so we were suspecting that already the DI releases they were planned beta one, beta two they were planned to be very close together and there was like three months between two so that was also already a signal and the RC one is nearly ready but certainly not for a three weeks schedule I think I tried to bring Otavio on ISC but he's not around so I make this comment on my own and I'm not the DI release manager I'm feeling quite unsafe on that topic sorry for the delay Mark on a more lightly subject if we are three weeks or a month or two months away from the release we will need a new name in three weeks, a month, two months do you have the name? before I was brought up to this talk this morning I was asked by Clint on ISC what the release name for Lenny is going to be so I can now tell you Clint that the release name for Lenny is Lenny he did ask for it unfortunately I have absolutely no idea this is one for the release managers in their shiny purple suits that lot they get to decide who the release managers are namely the ones at the right hand side and at the front that's a look in HE get to decide what the next one's going to be I know that at the previous FOSSTEM meeting we did have HE was asking for the release to be called RC Buggy but that does seem to have been linked to experimental now by our lovely FTP masters at the back especially Ghana for that who I'm sure is going to give you a nice big wave there's been various other ones talked about but honestly I have absolutely no idea what it's going to be I believe there was a slight issue with the etch release one of the reasons the current BTS command, BTS system doesn't recognise the tag Lenny ignore is I believe because we didn't tell anyone the name of the next release until after etch reshipped or something so there may be a problem with that Yeah so someone has to tell it on again I have absolutely no idea I'm afraid so anyone else? I don't know if you read my blog post of this week which I will try to summarise I feel that we are the release team tends to treat DDs bad by default which is quite strange for example you asked for reviews of everything that we want to get into Lenny because in other projects usually the social pressure of the other developers is enough even in big projects like GNOME the social pressure is enough to make DD to make developers do the right thing about getting things into in stable release I mean if in GNOME someone tries to push changes that are really clearly too much for the next release it just gets flamed and that's enough to prevent him from doing it and in Dembian what we do is we put a lot of load on the release team to review even minor packages that mostly nobody cares about and that's quite interesting that we can't solve this socially and we have to use a technical solution for that yep I certainly read the blog post I haven't had time to actually think about it much yet I was busy sticking hackagotchi heads on random photos instead of trying to prepare the rest of the talk as you can probably tell now all these issues mentioned here were found in one package that was Astroenum blog if it is I think this is it is possibly a social problem I don't necessarily agree that it's a technical hack around the social problem I do know a lot of release systems do involve code reviews and code freezes to try and make sure there's a period of stabilisation if we can get all maintainers paying more attention to the fact that only small changes and non disruptive changes are needed and that all our seat bugs are being fixed then that will make car life a lot easier at the moment I'm not sure it does exist is this something that can be solved easily I'm not sure certainly for the moment the freezes is something we've brought in quite a while ago to have a period of stabilisation if it can be solved by peer pressure that will certainly be interesting I'm not sure how well that scales with the amount of developers we have Mr question directly related to that what's the average percentage of the Act rejects that for propagation off the top of my head I'd say probably about 80% say get accepted so that basically goes directly at the whether or not we can remove the technical restrictions so there's 20% of people who don't quite know what should be put in or not so then the question is how do we work out what that 20% is without creating an unstable testing version right and also I mean how to avoid making changes is a good thing I mean we don't want to necessarily discourage that we just want people to realise that hey this change is destabilising upload to experimental or something I mean has anyone noticed that there's been any particular big delays with getting stuff unblocked it could only take a few days or even a week to get something unblocked which is not that much but it's not zero delay as well considering we have two years between releases right and also considering that so long as the unblock happens faster than the mandatory delay period in testing anyway or in unstable anyway it doesn't really make a difference so I mean considering most of the packages are uploaded with the 10 day delay then so long as it happens within 10 days it's fine also I should point out for larger packages we are now I think it's since the last release we now have a new hint called aged days so certainly with some larger packages which we normally wouldn't accept we can say set the aged days to 30 or so or any particular longer value which would help give them more stabilisation before they enter testing so that's also another possibility so maybe a difference between the social problem pressure in Davian and GNOME is that in GNOME every GNOME developer cares about the whole of GNOME more or less I don't think that many people in Davian would care for for example the proof assistant packages that I maintain in Davian they care what they think about is I don't know for example web applications and they have no literal interest in that kind of applications GNOME is less diversified again yes this is possibly a problem I think we are starting to see a larger move towards team maintained packages generally a larger team oriented way of developing things there seems to be a lot less emphasis of a particular package maintainers package is their domain and no one else can touch it there's been certainly quite a bit of emphasis on NMUs are not bad there are other people trying to help your problems and various wide ranging things which happen across the board so hopefully this will certainly help and people will become more aware of our overall goal which although some people may not agree with it is to actually well in my opinion is to actually release and get a release out to our users certainly one of my priorities anyway so we've got about 10 minutes left now so anyone any other questions or anything I guess not so thank you very much everyone