 Hi folks. I'm here with some other folks from home to talk about the state of the ARM ports in Debian. This is, as I've said, it's very much a boff. I'm not planning on talking all the way through. Pleas interrupt if you have questions. There'll be plenty of chance to discuss things. Let's go. So, I want to talk briefly about the existing ARM ports yn Debyn, y bydd y Bylydd y Hardwaire. Am 64 yw'r newis hyn sy'n gweithio'r gweithio'r gweithio, yn gyffredinol a fydd eich gwneud yn cwestiynau ar ôl. Mae'r bwysig yn gwneud o'n gobyn a'r gweithio'r gweithio'r gwneud y gweithio, ond o'r boffs ymau a'n ddechrau, rydw i'n ffwrdd sy'n gweithio'r gweithio'r gweithio'r gweithio'r gweithio'r gweithio. Felly, rwy'n meddwl, Ommiel yn y gweithio'r amser ar gael. Mae'r gweithio'r amser yn Lenni, yn y ddweud yn gweithio'r amser. Felly mae'r gweithio'r amser yn Debyn, ond mae'r gweithio'r amser yn gweithio'r amser yn gweithio'r amser. A hynny mae'n gweithio'n gweithio'r amser. Ommiel y gallai gael oherwydd y awddai cymaint, rydw i ni wneud i ddweud yn y dyfodol. Mae oedd rwy'n dod yn y gweithio'r amser fydd yn gweithio'r amser yn stoleim honno fwy oed, ac mae'r gweithio'r gwneud yn ddotr. Mae o waith i ni ddaeth eich gwas Blackened Pений. digital. The soft float, the EABI for V4T on processes the T means it has thumb. Some of our old hardware still wants this. It's supported upstream in kernel, in toolchain and a whole range of different places but that support is starting to go away. Various projects when you say that you support the still support V4T will point and laugh rymdynol am ni ddiogel yn ddylai. Mae ni'n eu vraag yn sylitudnol byd séig? Rwyf wedi gwelio'r ddaf yn ddymarfodol y ddyfannol mwy ydi, rwyf wediw y ddymennu. Be gynnydd i fynd y ddych mommyad. O. Fy fydd yn ychydig. O ni'n sylwydol, dyfannol mewn 7 i ddyl unlimited i'r mynd fod be gynnydd o'r cyfrifio'r cyfrif yma enghraifft. A'a gwertho'r gwertho'r gwertho'r gwertho'r gwertho'r gwertho'r maen nhw, mae'n dda, mae'n meddwl yn daw i'r bwysig fel mai'r llawer ond hi at pethau gwaith i yr hyn. Mae hyn yn oed yn dda i'r bod, yn oed. Ond ydych chi'n meddwl hefyd yn adrodd. Mae'n ganydd â'r mwynhau o'r dddorol o'r parir yng Nghymru i'r Debyg, ac ychydig am ymddefyn i'ch meddwl, ac mae'n ddiddell i'n meddwl, neu fod yn ei oedd ychydig o brinag, Yn ei bod yn ymdau i fan gael ffii yma i'r F7 a beth eich bod ni'n gael ffii, F5 mae'r ddweud yn cael diemio am gyda'r bwrdd. Felly, mae fyddwn yn cael ei ddweud nesaf o fwy oherwydd yr armiel yn gweithio'r gweithio. Mae'r ddweud o'r ddechrau i gael'r ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud o'r ddweud. I'm not proposing that we dump those people at all, but at some point we may move forward from V40 to V5. I'm forever being pushed by, say, the arm tool chain people to ask, oh my God. To ask, say, hello Colin, to ask why are we still doing this. They would like us to move forward, but whenever we ask can you give us a real reason why, there actually isn't that much difference between V40 and V5 in the instruction set. So anyway, that's a heads up. OHF is the other existing arm port that we had more than two weeks ago. We first released that with Weezy. This is for the much more recent. Does the mic gone away? It should still be there. Okay, it's there, but I guess somebody's turned it down. Am I okay on the stream? Yeah, it's definitely there. So OHF is the more up to date arm port. Again, it's EABI, but specifically it's the hard float variant. Targeting on V7 with VFPV3D16. I won't bother explaining that unless people really want me to. Just suffice to say, it's current hardware that is basically in your Android phone and is on almost all of the V7 devices that are shipping today. It's the standard that all of the ARM Linux distros basically agreed on a couple of years ago and then most of them stuck to. Don't get me started. I could rant all day and I don't want to. I'm a nice person. With the latest move over to the ARM multi-platform kernel and using DTBs, the nice thing is we no longer need to have three, four, half a dozen or a hundred different kernels to be able to support all of these devices. Not all of the supported devices yet have all of their onboard peripherals and whatever supported through a device tree. At this point, no new platforms are going in without device tree effort and anyone who hasn't shifted to device tree yet for ARM really, really needs slapping. Unless you've got good reasons, I don't know. I don't pretend to know all of this. The nice thing is with this that we don't necessarily have to go out and test on every single device. In theory, a well-supported upstream kernel port should just work with a mainline kernel and it should just work in Debian. We've got a potentially massive set of supported devices. So build these in hardware. The picture though is a bit small, I'm afraid. We are mostly still relying on dev boards for building RME L and RME HF. We did have some older machines, but the current stuff is Marvell were very, very nice to us and donated a set of eight on modern XP machines not that long ago. Those are quad core. They have four gigs of RAM on board. They have on board SATA, four lots of giggy. These are really, really nice machines. Whereas on the older IMX53s we had, it could take several days. We had a little mini rack full of these and we were still not very fast. Each one of these on modern XP is basically about as fast as maybe five of the old machines. This is why we really are screaming ahead. The older V5 buildies that we were using for RME L in fact were also Marvell donated dev boards, but they were so old that actually they promised us they would donate the new machines if we promised to turn the old ones off. We were approximately the only people on the planet who were still using the kernel support for those old machines and they didn't want to have to maintain them any more. So here, have new machines. It's nice. It's good of them. We can now drop the kernel flavor absolutely. There were some bits and pieces. There were some teathing troubles with these machines. Like a lot of experience that we have on the more embedded platforms that we do endeavoured, the more embedded ports, the hardware that you get available that you get donated isn't actually expected to be running 24-7 churning with all the cores and all the memory and all the disks. We had quite horrendous heat problems with those new machines until Vince and I spent quite a lot of time physically modifying the cases and putting bigger fans in. Imagine a typical gamer PC. We did that kind of thing to these boards, physical hardware mods to the cases to make them go cooler. They work flawlessly. So thank you for your help with that, Vince. We keep on having offers or we keep on chasing offers of newer hardware and real servers and told, I'm looking, but I'm not getting much response, we did have offers of wheel-on-based servers, but it not quite happened yet. I'm hoping it'll happen soon, yes? It's actually less of a critical path now that we have better machines already. But still, wheel hardware, wheel servers that are designed to be managed and run 24-7 is better. So, moving on. ARM64 is the brand-new port that went into unstable exactly two weeks ago today. I'm not going to claim most of the credit for this. Wookie is sitting down for some reason instead of standing up next to me. Well, come on. Wookie has done, frankly, almost all of the legwork to get ARM64 built in Debian ports. So with all of the massive amounts of the usual thing of bootstrapping and breaking all the loops and everything, you've been at this for 18 months? Two years? Three years? Yeah, depending on how you count it. So I now get to do the easy bit of saying, well, when would bootstrapping ARM64 into the main archive? Oh, we have a build loop. I'll just go and copy a package from the Debian ports archive and then that works. He didn't have that option. He had to go through by hand and work out. So what do I do to bootstrap and break this loop? And steal bits from Ubuntu? And steal bits from Ubuntu in cases. Two years ago? Yes. So we're now in the main archive. As of this morning, we were just about 50% of the archive built. We've got plans to release with Jesse. The release team and all the other teams involved are happy with that. They assure me that they are. There are concerns which are not wonderful, but to be honest, they're trusting us. We trust them. We're nice people and they trust us to make it all happen. We have two official buildies right now hosted at ARM, which are the ARM Juno dev platform. Considering again these are dev boards, these are six core with eight gigs of memory, USB disks because that's all they support at the moment. But wow, they've been really, really stable and reliable and really, really quite fast considering these are dev boards. So we bootstrapped using just two machines from nothing to 50% of the archive in two weeks. That's not bad. You probably can't see. Apologies. This looked better on a small screen. It really does. If you have a look at the two really, really steep lines on the right-hand side of that graph, you'll see that there are ARM64 and PPC64EL. We both went into the archive at about the same time, a few hours apart or whatever. We started first. The PPC64 guys came along with their massive server hardware when they overtook us. That's cheating. Of course, it's not a race. It's not a race against each other. It's a race to get, frankly, enough of the archive built by the time that the freeze starts to happen in about two weeks' time. But we want to make sure we have enough in that people are happy. Colin, pass the mic. Why the tail-off? The tail-off is, I'll explain, Colin probably knows this already. The tail-off is basically the very first set of packages that you upload tend to be quite small library packages that don't have lots of dependencies. As you get further on, you'll see PPC64 is also tailing off, not quite as rapidly. That's why as you get further up the dependency stack you have things that mean you need to install a lot more build dependencies. That takes time. Secondly, they're the bigger packages. It's not clear to me why the things aren't the same shape. Because we're both doing the same thing. It seems to me that although one machine is faster than the other you should still have the same shape and we don't. On average, yes, they should be then not quite. It's not a blockage. Both those machines, I think, have been building solidly since they started. There's no gap. If you look at the Debian ports thing there were some blockaging gaps and pauses in that. You get to about 70% and then things really slow down because then you get onto really fat packages which take forever. There's a reason that when you have plenty of package built you have also more chance to have a package, a new version of a package. You have to build it in addition to the bootstrap itself. We did also, there was a slight screw-up as we were bootstrapping we didn't have current pull. We unblocked a whole load of pull packages and then had to do a pull transition which then meant about 200 more wee builds. They were only taking three or four minutes each on average which is really nice on new fast hardware but still that's still 600 minutes of build time on these machines. Neil, as we've been going through we've been explicitly, there's a whole slew of us on the hashtag being buildy. I've been watching the needs build queue, the build depth on installable list and whatever and making sure we keep on trying to get the needs build list as high as we can. Last time I looked it was like 1,200 packages. The last thing we want to do is have machines going idle because there's nothing they can build when we're in a race. As I said, not against PPC64 in anything but whatever they've got flashy hardware the race is against the deadline for Jesse Inclusion. I've been following a number of dependency chains just trying to keep an eye on some of the leaf packages some of the ones that I am actually maintaining but they are at the top level of dependency chains so it's very easy to track which ones can't be built which ones are being blocked halfway along. The difference in the PPC64 EL line and the R64 line I think is actually because the PPC guys have chosen a different path through the dependency breakages and a lot of the packages that we've now built on R64 and which are taking out of time on the Junos haven't yet been built on the PPC64 EL as well. They've got a different path through and they're still doing smaller packages. So the two should converge eventually? Yes of course. As we get up to 80-85% we're going to be converging guaranteed. As Wookie will attest and he's got a talk coming up this afternoon on the next talk. Bootswrapping is fun. Bootswrapping is bloody hard and really tedious. It's harder than it should be. The things that you find, again, join hashed ebbyn build D especially if you go and have a look at history if you can find it you see the stupid things that mean that typically documentation means that things all depend on Doxygen or TechLive or whatever. You end up with silly loops where GTK3 and QT4 end up stuck in the same loop and you end up with packages that you just cannot understand how possibly they could ever make sense for them to be in the same build chain but they are. TechLive needs GDB. They're really cool things like that. It's hilarious. Anyway, ARM64 is really cool and it's coming. So talk me of ARM64 and he isn't here right now because I think he's trying not to die somewhere in his room. He got the debcon for Lergy and hopefully he won't die because he's a good friend and I don't want to have to pretend and know how to do hardware. He was meant to be doing a talk later today about a project that several of us are wanting to get off the ground which is developing a laptop using an ARMv8 an ARM64 system on chip. The particular one we're looking at to start with for prototypes is basically a replacement motherboard to go in, I think Pad X220. So you'll be able to use all of the nice bits that come with that machine but with an ARM64 and lots of memory and whatever inside. That's a project. The actual talk has been moved from this evening because it clashes with somebody rather more famous who's turning up. I think people said Andy could still have his slot then. He wanted to actually have somebody to talk to. I think it's meant to be happening tomorrow morning. I don't know exactly, you'll have to check the schedule. I heard tomorrow at 11 even so it's worth checking in the schedule. We also have for you today we have some donated hardware. Steve at the back I think has a box full a box with a few banana pie boards in. They spoke to Martin Mickelmeier and Steve and I'll let him explain a bit more. You've already covered the salient points. I have hardware and it's free. We have a small number of deaf boards that have been donated by Remaker. Martin Mickelmeier had these shipped to us for giveaways at Devconf. There's hardware if people are interested in bringing these boards up running Devian interested in improving the Devian board. He's asking me if I want one. I'm saying I have all that free time that I might spend on it. If you think you'd be interested in having a piece of Arm V7 This is V7. This will run Arm V8 Jeff. Imagine it's like a Raspberry Pi but a bit more advanced and it will run Arm V8 It's like a Raspberry Pi but with bananas I guess. These are banana pie boards. The chip is the Allwinner A20 It's an Arm V7 chip We have these donated to give away if you're interested. Anyone, hands up? Come see me. It doesn't work with a Devian Colonel yet. It probably won't be long I hope. If you're interested in putting your hands up and keeping them up long enough for me to actually see you. We have more than enough people to be honest. If you come and grab me at the end and we'll talk about it obviously we would love to give these out to the people who have the time and whatever to get these going and equally especially the people who can't just afford to go and buy one of these themselves just because. If you think you're a good fit, please talk to us. The other thing that we have to talk about today is something that Pavel, my colleague's mom will tell us about right now. There's been a post-op talking about this. It's donated software. I'll just use it in the old school way. Actually the Allwinner A20 It's a nice little chip. It's got a jewel core. Sorry? I'm trying to but it doesn't work well with the glasses. May I be like this? Cool. So the A20, the Allwinner A20 is actually a nice little chip. It's a jewel core A7. A7 is probably my favorite V7 core. Very small and quite nifty as well. Anyway, there's something else I wanted to talk about. So I work at DSG Development Solution Group and we are the part of arm charged with providing tools Development Solutions or if you wish. We do both commercial and open source tools. We've got probably about 20 people working on new tools both the GDC, Binutils GDB as well. We've got 20 engineers working on LLVM and we're still hiring by the way. But the topic of today is DSG Development Studio. So it's a commercial offering normally. It's an integrated tool chain. It's a debugger and a profiler. You'll see it in a second. The important thing is it works with both 32-bit and 64-bit targets. So both V7 and V8. And it's available now for free of charge for all depend developers. And yes, it's a proprietary software so I'm sure that there will be people who won't touch it even with a stick and I respect it. However there are some people who who will use whatever it takes to get to the point and I know what's the background of this particular picture. I just wanted to recall this small part of current development happening in BeatKeeper and something good came out of it. We know all the revision control systems actually got much better once this gentleman started working with them. Anyway, so I don't want to make this talk any more about the marketing that's really necessary. So there are two main parts of DS5, the debugger which gives you pretty much everything you can expect from a debugger. It will connect to a standard gtb server running on the target. This bit is even niftier or rather nicer. It ticks all the marketing boxes. It's colourful, it's simple to use allegedly etc etc. I'll show you this live in a second. What to do in order to get hold of this tool. Go to ds.arn.com slash Debian website. There will be a small, there will be links to downloads with the installer of this product. I'll talk about the installer in more details in a second. In the RNG or brownie box there is a place for your Debian developer login. If you type it, you should see something along these lines. So there will be an acknowledgement of your request in some hopefully short time you will get an email encrypted with your GPG key containing a serial number a magic number then you then type in a particular price in the product and it suddenly all works unless the licensing is broken again. So a quick how to. The current thing that you will find on the website is a installer in a form of self executable script it's not packaged in the package yet this has been on the packaging for both Ubuntu and Debian has been on files for ages. It's just not enough critical mass for us to actually convince validation people to start and build people to start building this kind of distribution so the more the merrier the more people are actually using it from the Debian community there is a greater chance for this to happen. No root is required it will by default install in your home directory alternatively you can install it as a plug into your existing Eclipse installation it's entirely up to you. For the streamlines so for this colorful thingy you've seen generally speaking you want to install a out of tree kernel which is shipped, it's all GPL it's shipped with the product and a user space demon as well these days actually you can see it in a second you can actually make it work without this out of the tree kernel so it's in the simplest possible scenario it's just about copying one binary to the target running it there it's a demon exposing it's cp socket and the streamline will connect to it now I'm sure there will be issues all I can ask is to bear with us it's a new experience for everyone let me just say that our product product people, the business people for DS5 weren't entirely comfortable with the idea of giving away money and when I got a bunch of serial numbers I was told that I just hold something like 100,000 bucks in serial numbers don't believe in all this stuff it's for the good cause anyway so if there are any issues if you have any issues whether with the formal side of requesting license or with any issue with the product feel free to join the tools group on the Arm Connected Committee is a web forum or contact me or any other of the Arm guys around here and that's all for the slides and if we still have a second I will just connect to the target which I've got here it's a it's a big old bone black wired up to my obviously running Debian and what I will do is I'll just start capturing data what you can see now is a live view of CPU usage some data from the performance monitoring so for example there is a there is a chart showing the instruction being executed so if I just quickly go to the target and run some meaningful load on the system that's my favorite well it's true semi-colonious in the world plays yep ok we shall see more activity hopefully and once this is done there are some other features so you can now see at the bottom you can see the processes and thread in the system obviously the color of those of those slices means that the process is hotter or warmer etc etc and there is a flat profile if I've included all the debug symbols for the running process I will see some more detailed information about the code being executed you will notice that there are only two tabs in this window it's because I'm not a Debian developer therefore I have no license for it therefore you will see more having said that I use Debian from something like 1995 so I probably predate quite a few of real Debian developer these days anyway so is all from me feel free to catch me on or question so Debian has eclipse builds for army hf and arm64 doesn't run natively on these platforms streamline will work I've tested it myself streamline will work the debugger I do not know having said that one of the reasons I've managed to convince our product people to actually do this is that it is supposed to be the first step in the direction of native tooling so we are getting exactly where what you've just mentioned we are getting to a native tool so self-hosted debug and self-hosted profiling yes I say streamline is tested if the day was longer slightly longer 24 hours the debugger would be on the list as well I have quite a few products right now and I'm actually working on the gator so the streamline agent packaging as well more question but I will move I will move away feel free to catch me at any time later on I'm around why the cables always go the one way around hello are we there sorry I will be with you in a moment yes clearly anyway that was the main point of that I'm basically done with the bits I wanted to say and we do have some time clearly this is meant to be a boss so please let's talk about stuff Dimitri hello I believe last year there were some talks about bringing rasbin into the archive what's the state of rasbin very isolationist in my experience the rasbin guys as with most projects they don't have a lot of spare time so they seem to be very isolationist they'll do what they need to do to get what's good for raspberry pi done and then that's all they're interested in so we talked about it after the thing and if we'd started there they would have joined in but as we haven't and now it's all working they couldn't be bothered changing everything just to be part of debut because it doesn't really change anything about what's available to people it would just be work so I think it's reasonable to say they couldn't be bothered yeah one of the main guys behind rasbin is a Debian developer and he's super active with derivatives I think I'm not sure and I'm roughly involved in that idle and lurk in the IRC channel but I think they'd be receptive I don't think we want another arm architecture for v6 plus hard float when we have our ahf especially now we can get a selection of pyre likes which are in fact v7 boards so things are moving along let's say something I mean it's already the point that we're adding on 64 and we're taking flack from DSA the release team and a number of folks not unreasonably to say do you really need three different ports in the archive the thought of doing a fourth or fifth or more let's not go though it's diminishing returns we could potentially move RML and rasbin and merge a bit I mean to be honest if we moved RML4 to v5 it's still not going to use the hard float stuff which in theory should help at least a bit for rasbin on the v6 to be honest it's a shame that the particular GPU CPU combo in the pyre happened to be v6 if it was v7 this would all be much easier but hell that's the board combo SOC they went with so what's the future of the RML port asking because for example for full C++ compliance we need the atomic helpers which are currently not available for RML4 and 5 it's not an issue I think for C++ but it might be an issue for jesse plus one the helpers are all though in the kernel they're just not going to go very fast that's the thing the stuff that has come in for C++ 11 absolutely depends on features that on v4 on v5 just don't have so you need to go into kernel it means performance is going to suck to be honest if you're building a large C++ thing you have swap right yes swap is that in v6 but swap works on v5 you can emulate all this stuff so I don't see why you cannot do that sensible is a different question of course Ben do you want to pass the mic over if I understood rightly the atomic helpers provided by the kernel for old arm they run in user mode but the kernel knows exactly where they are and it can do special things if they're preempted and that's so then they're probably more expensive than the native atomic operations of later arms but they're not that expensive yes they're helpers they're not cest calls they still hurt it's nowhere near as fast as if there was real native support for these exactly they can be made to work I thought by now people already had them working it may just not have filtered through yet fine I mean again it's for the newer stuff that's using them I guess we'll need to work out how much effort it is to continue for the people who continue to want to run RML say for their dream plugs or whatever they may not care about newer stuff that's using this we may possibly get on a limited set we might move forward who knows it's a discussion to have on the list if people don't care but we need it in the archive to for example fulfilable dependencies then we don't have a complete port anymore so maybe Armial will be the first partial port maybe that would probably be sensible too people are not going to be running at least I assume because I can't imagine how slow it would be you're not going to be running ice weasel on a goer plug or something are you going to run out sorry you're going to run out and ask for it in C++ yes if you say the arm tools the tools will no longer support VIV 405 what for with the use of these tools for non-linux targets as across development environment because there's still quite a lot of arms arm 7s out there and people still want to write code for them which is he they have to run full Debian but some small artels or whatever but they still want GCC so we'd be surprised if GCC would decide to drop our 7s I can't imagine the GCC folks who are about to drop arm VIV support anytime soon as Pavel said we have a whole load of people working at Arm whose job it is to make sure that GCC works well on the old stuff and the new stuff the IMX 28 which has been released Mike the free scale IMX 28 which has been released probably two years ago is still based on arm 926 because it's small and it's cheap so this will carry and the tools will definitely as in the compilers and the likes will not drop the support for VIV 40 or even before but definitely for Debian purposes moving forward Arm HF and Arm 64 are the things that we would expect most people are going to use so we're always looking for more help with porting like all of the ports you know where to find us hash Debian Arm or the Debian Arm mailing list another thing that I have mentioned on Debian Development Outs fact this week we're having a mini debconf hosted by Arm and Cambridge in November this year there will obviously be quite a lot of people who are interested in Arm and Arm related issues that's not all that's going to happen it will be a generic conference as well so there will be a lot of other issues and a lot of other talks if you're interested please go and have a look at the wiki and sign up or ask me about it and we can talk again later any more questions, comments I could just say in case people didn't know we do have a bare metal cross toolchain in the archive since a few months ago which for all those tiddly arms that ours isn't proper Linux so we do have a hardware giveaway still of those banana pies wiki has some hardware to give away as well if he's remembered but he's not listening yes you do stickers stickers oh yeah hahahaha um I've got stickers yes you may notice there are some Debian powered by Arm things like we have on our laptops but hey they're free stickers if you're interested we have some of those to give away they were meant to be here right now but it seems they're not blame wiki sorry basically I've been putting some effort into getting a few armhf boards that are moving towards working out of the box with Debian installer getting flash kernel support making sure modules are enabled in the kernel it'd be really great to see a few boards that could just work out of the box and I would love to talk with people who want to help with that I will see if I can get Chetan Tiki on to work we had another hand near the back wow we rapidly wanting out of time so so when you have the stickers where will they be available hahahaha I'll put them on the front bench thing I've been carrying them around all week and I forgot today obviously I've been quite busy so definitely if you're interested in one of the banana pies please come and talk to us straight at the end right now and I think that's a good point to say thanks everyone I'm working for Arnold seconded into Lenovo so is Wookie there is a huge amount of corporate stuff pushing on Debian work at the moment and we're really quite happy about that they are doing quite a good job we can't be too rude about them really knowing the legal department obviously Wookie has basically been paid and it has been his day job to get the arm 64 putt going as well and that is really really awesome lots and lots I mean the reason Pavel is here today it's not just the arm you know want to get the want to get proprietary tools out of the community the reason for this is a massive number of the day-to-day engineers inside arm are really cool free software people love Debian love Ubuntu love all of the free software ecosystem and want to get involved we're going to be hopefully several more of our co-workers in arm there's at least two in our office who are wanting to get into the new maintainer process very soon so we've got some really good engineers we're slowly taking over it's working quite well we're taking over inside arm and they're trying to do a reverse takeover this way as well it's cool it's good fun so thank you very much everybody