 Thank you very much, it's great to see so many old faces here. I'll start with that intro there. And any new faces, if anybody doesn't know me, my name is Theodorat and I'm telling... here to tell you that pledge was all a big joke. Okay, um anybody been to my talks before? Yes, good good good good well this is the first one, ddweud yma. Mae'r un i'r ddullion arall, ond mae'r ddweud yn gweithio. Mae'r ddweud yn ei wneud. Mae'r ddweud. Mae'r ddweud... ddweud. Mae'n ddweud. Mae'n ddweud, maen nhw'n ddweud yma'n ddweud a'n ddweud yma'n ddweud yma'n ddweud. Ddweud. Mae'n ddweud. Mae'n gweithioulle, ein ichern, dat grammar 2000, y mggeith yng Nghymru a mae'n gweithio Constwctoral. Mae'r ddweud. Mae'r ddweud. Mae'n dod odd. Mae pawb hyn yn hyn mathematics. Mae'r ddweud.bent y myth o'r NETBSD-8. Yn gweithio, yna gweithio'r ystafell o'r hynny. Fy fyddai'r bethau gyllidau cyffredinol. Rydych chi'n gweithio, yna hefyd o'r nesaf o'r ffawr? Yn ddwy'r gweithio, mae'n fawr ydw i'r lleol. Felly yn gweithio'r NETBSD-8 o'r ffawr? Fals? True? Oh, we got some truths! Yes, excellent! Right! You need to keep chatting them out, folks. Right! Um, well. I've just realized that presenter notes aren't coming up here, so we're going to have to go and see. Um, 7.0 released in September 2015. And so, we're talking about roughly two years between seven and eight. That's longer than we like to have.giggles ideally it would be a year but it looks like we are going to ramp up that and aim for a year going forward as well so we expect nine to be around about this time next year that's the plan anyway we've we started the process in June 6 it's still there in the branch, it's in beta at the moment and we expect it to be released real soon now rwg, i'r bwg ac yn ymddangos ar y dyfodol ar y cyfanydd iawn. Rwy'n cael ei gwaith. Rwy'n cael ei gwaith. Rwy'n cael ei gwaith ymddangos. Dwi'n cwer o'r ffalsu. Ffalsu? Ffalsu, ddwi'n cwrw. Rwy'n cael ei gwaith. Rwy'n cael ei gwaith. Rwy'n cael ei gwaith. Gwyd gys. Ond nid. Ac mae'n rhan ddod, rhan oedd yn ddod. Rwy'n ddod. Rydych chi? Rwy'n ddod. Gwybod, Sabamau? Rwy'n ddod. Rwy'n ddod. Yn ystod hy, mae yn hy, hy, hy, hy. Yn ystod, rydd, rydd. Hardware Hardware So we're going to be talking about hardware here Right, next one Old hardware only NetBSD False Any truths, any truths? No, no, no, nobody's going to stick their neck out Okay Old hardware only Well Gerard McNeill up in Newfoundland has been quietly overachieving for the last year and a bit, I think All of this is his work with the ARM ports in NetBSD I don't know if any of you has taken the time out to compile the EVB ARM port just recently I did it just after an AMD64 build on a slow VM which took five hours The EVB ARM port took 11 hours and that's mainly because of all the kernels that we have there We are killing some We're killing this in huge detail So we're talking about old hardware The kind of Cortex names down the side you probably don't realise You don't know But the ones across here are fairly obvious, very well known I think in the outside world So we're not just there on old hardware We're there on new hardware, especially in the ARM space Now Gerard explained all of this over about an hour to me and if anybody wants to go and have a look at the type script in which he did this it's very, very instructional I can tell you that It's in the form of a script file so you just play it back with script minus p that and that should work on all the scripts that are out there in the BSDs But it gives a great idea of what he did why how he did it and not just the hardware stuff as well all the full device trees and other things like that that are coming on So Intel QAT support came in from two of the IIJ developers I'm not sure if it's in Tree at the moment but it is there and the ultimate support for new hardware there and there's actually a screenshot of this I don't know if you can see down here that's the frame buffer and where is it on there maybe it's the next one yeah I think that's it there USB there at TTI brilliant Felix Dortmund in Germany did that and that's marvellous So any guesses true, false partial what way partial I'm compared to GCC we don't have all the architectures that they have for example but there is support for other newer architectures there as well so yeah okay I'll give you the partial there Chris okay it's in Tree, Jorg has been maintaining it it's still optional and it's off by default Jorg pointed out that the regular testing with the bulk builds is particularly good at finding compiler regressions and the other thing I'd note is that on the bulk build stats that we see on the mailing lists the clang bulk builds typically have way more packages building not way more they have more packages building than GCC builds whether that's because Jorg is more efficient at doing these things or whether Manuel doesn't do them as often I'm not too sure there we go if you were in the talk before this one Camille was talking about LLDB and the debugging infrastructure one of the particular points that Cherry made there was about the the libfuzzer and sayfield that was there and the sanitizers as well are very interesting for us and Camille's been doing some great work there was that okay okay true, false maybe partial partial because we have no top of the kernel virtualization okay we've got Zen of course had that for well over 12 years now user mode it doesn't work actually if you compile it just recently something in the changes that will be actually sanded for the LLD scripts have made it not to compile we've got Rump which is the kind of ultimate virtualization by namespace munging I think is the best term of it but we have no beehive or KVM equivalent or even the the open BSD BMM that's probably a big hole that we have be great if somebody would have a look at it okay I can make a few signing I can make a few signs nope I think Rheostrad is trying to give me the American Sign Language for something there but I'm not sure which it is any guesses partial we're talking about net BSD here so I'd probably go for a no we don't have it but we do sign hashes of all the releases that we do which is a minor thing it's not that great and you'd have to generate the hashes yourself to work out whether it's good or not but having said that all of the metalog hashes that are created during build.shell they get installed on the machines so it could be done automatically we do need to pick up a game here though right outdated utilities in base and Christos isn't allowed to answer this one true false what sorry I've got a script that will help you out there Martin but colour LS is in package source and I have a script that will do it because I need colours I need all the context I can get with my eyes anyway I'll say a no Christos is doing a marvellous job there updating all of the the third party stuff that we have in base there is now a small shell script called sysinfo which will do a fairly fine grained job of working out what specific versions of packages what specific versions of system information that we have in the base system and as of last week I committed some stuff to do it on a dester as well so we can actually work out from the dester in build.shell what packages keep saying packages yeah that's the next step yeah small steps lots of them it's coming sorry ok just to repeat the question Christos said he wanted it to be extended so that the latest version could be printed relative to this this is actually a snapshot of what's installed on the system and things like that it would be nice to have that up on the website and so people can actually run it locally and check it against of what's there going to be the next stage of this whole thing ok I'll let you off without answering the previous one partial I would say probably right we don't have any of the new modern congestion control algorithms or mitigation algorithms in our TCP one of the bright spots was manual buyers SOC can implementation for all of us who have cars and want to run NetBSD in our cars and boats sorry my mind was on other things yeah and the only other thing I could think of was that Roy's done some marvellous stuff with DHCP CD and I used to think that stuff I'll just use DHCL and it works fine but if you actually have a look at the size of the executables the DHCL one is about 5 meg and DHCP CD comes in about 60k or so so great work and I know Christos has been working closely with them on getting some bugs fixed and Roy's very responsive and it's great to have that but I'd agree in general networking is outdated so is it difficult to develop for false thank you yes we have built up Shell it's been around 15 years does cross OS building and cross platform building that's incredibly useful and a great selling point as well for NetBSD we've got the make image there which will build up an image for anybody who wants to run things and especially nice is that you can actually add packages into that and get a straight image built for you like that it's now runs on privileged and we've got other things in package source that will make live CDs and make a USB stick that you can boot from and just recently Jared's done some marvellous work with all the U-boot packages and Savannah as well yes there sorry didn't mean to wake you up and mentioned all the U-boot packages that's the ones I could find down there that would fit on a thing apologies you probably can't see them very well but yeah it's because there are a fair number of them these days okay run by morons well well I have to say in my case it's absolutely true if I have anything to do with it others take a large amount of time and effort to run the the project and to do a whole lot of the administration of the Petra of the bank for example we have a board of directors and a core team which is the technical management so it's just like a company set up really we have AGMs every year and board elections every year as well so this is your chance to become a moron right how about this one true false I'd have to say false actually we have no hammer one of the reasons for that is that it involves VFS changes our ZFS is out of date but Chuck's in the process of updating it looking forward to that LFS was worked on during the Google summer of code and Puffs and Pood allow a few file systems reliably so for all of the people who said that we don't have modern file systems what ones are you thinking of okay yeah I'll give you that right so ZFS what are the other ones that people are looking at EXT4 NFS4 that's true I'll give you that what's the state of the art on Linux at the moment have they given up on B3FS XFS is years old and doesn't work reliably I believe the comment from the audience of the livestream was that Red have pulled the plug on BTRFS so yeah so right okay so there's still some discussion about this but we're not doing that badly we have the file system switch that all of our backups are done using on TNF machines so we're not doing that badly I don't think we've got SAMBAFS and the client one there true true no multipathing but yeah okay so maybe there is room for improvement maybe I was over optimistic or something like that I thought you'd added a new AES okay it's probably fairly constant on the machines I run it on anyway no detrace partial again because it's old sorry true yes yes and we certainly don't have some of the mods that George was talking about earlier on this afternoon sorry stability yeah okay I'll give you all of that however Sivan has done the necessary and added in the detrace tool kit scripts so it's not forgotten about and we'll certainly track the open detrace stuff when it comes into well out of github I think but there I see somebody here was doing stuff in 2015 yeah no NVME support false thank you yes we brought it in Nonakasan and Jeremy Dolcek brought it in from OpenUSD and it seems to be working fine certainly with his test okay false okay we've got a false and a partial and no it's still out there for review and he's still making changes I saw some at the end of last week but there's definitely hope in sight for this kind of thing okay developers out of touch for me obviously true right whenever I travel I don't take keys with me I have loaner equipment and things like that means nobody can email me it's great it really is if you want to relax for a while then do that so for my part I'm out of touch when I'm here when I travel yep did everybody bring their private keys and two factor tokens no no okay so I'm going to claim claim victory on that one I think okay next one you're very quiet out there come on no I'd say that we we do have them Yorg has been beavering away on this for is it five six years you told me this morning yep okay he's been doing conversions from CVS through fossil on onto git for five years or so there's been various speed ups along the way and things like that but we certainly have git mirrors of the CVS tree the CVS tree is obviously the master one and that we've also got the bit bucket mercurial tree there which is fairly new last month or so and that's great for all of us who like mercurial and yeah thanks to Yorg for setting that up and iig's we've got a repository that they do automatic conversions on using a fairly interesting javascript script so yeah that's fun okay testing is sorry i think looking at the project as a whole how would you see netbsd's testing in the whole scheme of things i think the combination of Rump and Anita and things like that give us a marvellous platform really for doing all of the continuous testing for working out whether networking has been caused to take more time due to one change and there's usually a list of commits that have happened between the two test times i think that's great so that's the rel-edge list that i think it was Christos was talking about this morning the platforms where the testing is done and the frequency of those tests so i think our testing is pretty damn good actually you're right it could be better we could test more stuff and i think that's the idea going forward as well in that we have a rule that any new utility that comes on board we need tests for it we need to be ATF tests now that's fairly onerous and things like that but we can always help you out with making tests into ATF tests some of us have done it a number of times and forgotten how to do it and then have to go back again and so we've documented some of the moves along the way i've also got some scripts to help us out doing this as well as well as that these are the runs from September 2017 August 2017 and we can see that the times and the dates where things started failing and the tests that are failing i think that's a great resource to have certainly in the run up to to a release if you want to find out if any people can remember back to the five and the six days we were taking six seven months to try to find out where something had gone wrong over the last two years this is a great way of helping us out so we don't do that all again and if you're in Christos' talk yesterday you'll know all about reproducible bills so yeah I'll send you a a redirect to the talk that he gave and the paper as well one thing I have to say it was way more complicated than I expected the number of things that had to be changed were marvellous were extensive let's put it that way I talked to Christos about it he said we were the first BSD to complete the Debian run and you thought they were the first big unix to complete the Debian run as well does that include all the linuxes? interesting rather a significant piece of work there and just a reminder that minix uses an FBSD user land in the build scheme as well so they've got this almost for free I imagine they'll have to do some changes but they're doing fairly well with that thank you we have now FDT in tree for those of you who aren't aware of that it's the full device tree which means that we don't have to specify everything in kernel config files example for EVB ARM boards and things like that it's a marvellous way of reducing the number of kernels and hopefully taking that build that I had from 11 hours down to a much more manageable 5 or 6 something like that so everything else that goes into internet BSD 8 the p-trace fixes from Camille that he was talking about at the end of his talk there there's a new sound driver it went in we had some issues but now everybody seems to be reporting that we've got good sound I think across the board so that's good Cytosans fixes for the faster Intel Nix and we mentioned the new AS mode constant time but I'm sure that's going to come next week or something like that okay one of the things that Jared talked about in that script that I pasted up there earlier was a new SDMMC driver which allows us to run much faster on the SD cards than we were otherwise able to Maxine Viards has done a PMAP overhaul AMD64 and i386 Marvel's stuff as well and Zencernal module that's the American spelling of it and there have been some changes to SCSI control Llewer is now at 534 in kernel which is great if you missed I think it was BSD CAN a marvelous talk there about a Llewer based security model these aren't eximerins or anything like that that was created by a guy not related to the project at all very interesting talk and it allows us to sandbox various applications and utilities just through a different security model so EXT4 FS has been beefed up a bit it's still called EXT2 FS in the file system but there are other bits in there it is, yeah, that's right sorry about that and the turbo channel USB stuff we saw earlier on all marvellous things the NB7 network changes are entry they're not switched on but they are coming so we're getting there and Nick merged his his USB work that he's been doing I think over the last two years or something like that again, that gives us some USB 3 support so we're a lot better than we were previously this helps us work on newer hardware and things like that GPD support including booting that comes from John Nemeth I think and the other marvellous forward looking hardware kind of thing that we have is PCI support for the for the sharks anybody still running sharks by the way well, excellent too I threw mine out a couple of years ago that was the last one I had but well done the pioneers here ok we brought in IP6 Adder control from FreeBSD Christos is Blacklist is new in NetBSD8 as well I think and Maxim VR wrote a scanner and has been running our source tree through looking for various leaks lock contentions and other things like that so there's a lot of security work that's been going on like that he also split the kernel P map and this is where I get into the section where I know nothing about it and so my mouth is moving and words are coming out but they don't mean anything but he's been doing some security work to to separate the user and the kernel mappings I think that are there I shall probably shut up about that now it's probably best if I do that Vi O Skazi which helps us much was a lot of the the virtualization platforms Benny brought us up on the computer engine so that's supported in 8 I believe and all bugs fixed and things like that and Riastrad has done some work on Nuvo I think it does so it's there now it wasn't before and we're good right so apart from having lots of slides here it looks like I'm going to finish early so I'll just say there's lots new in that BSD8 I've kind of clipped the surface of it here I think lots of people are working very very hard it will be available real real soon now for some version of real soon now and with that any questions and answers volunteers I was wondering if you could say something about the development of this particular release in comparison to the other ones has it been more painful or development you mean branching of it or do you mean the whole efforts leading up to the branch the effort so like the features and stuff like that so changes that cause problems and how volatile has the branch been from a personal point of view I haven't found it very volatile as in when I want to build it it's there previously I mean 10-20 years ago I come along and say what state is current in can I install it or something like that I haven't had to do that much etologist recently it's been there and I think the automatic testing the Anita stuff and especially the automatic runs have been largely to largely to blame for that lack of volatility or lack of instability I think I'd say we had some pretty nasty issues with the GCC update and we are not completely over that one yet but sure and well I think they were even worse in the past I remember M68K managed to take an absolute dive some time when we upgraded GCC purely because GCC weren't interested in M68K code generation and we had horrendous bugs like that but the ones and we're still seeing the fall out from that I think but I don't remember seeing any disastrous things happen in the last things I remember 20 years ago Mycroff managed to check something unless the tree broken for two weeks and refused to change it either because it was Mycroff I think the big broken things were all those subtle issues that didn't show up until you were just doing the white thing under a full moon yeah that's right my question is about the tool chain so I had a glimpse of some conversations about GCC deprecating some platforms recently I'm not completely sure I understand all of it but is that going to affect us and what do we you know are we as portable as we used to to different compiler tool chains? Let me try to answer that we have not upgraded to a version yet where series loss occurs for our architectures we are upgrading GCC in current right now to a version that is not affected by that change so we have the most longest time for that old platforms and we are actively trying to get those platforms supported by LLVM and even GCC the plan is not finalised yet as far as I can tell so well thank you very much for letting me be a stunt Mark I'll tell him