Jan 29 15:03:54 ==> starting "Meet the Anaconda Team" Jan 29 15:04:26 * jsmith-mobile has quit (Ping timeout: 265 seconds) Jan 29 15:04:47 while chris is working on the projector, getting started Jan 29 15:05:00 to start with, all of the anaconda contribs are here at FUDCon this weekend Jan 29 15:05:07 * listing off contribs Jan 29 15:05:38 * rdieter has quit (Ping timeout: 240 seconds) Jan 29 15:05:50 there is a spins talk right now, one of the contribs is there to give the anaconda team coverage Jan 29 15:06:08 basically, the anaconda team wants to talk about where they are heading and maybe get some input from the community Jan 29 15:06:36 the anaconda codebase has changed a lot over time, rewrote the storage code (maybe F10, no agreement) Jan 29 15:07:04 while this change was really good, it wasn't visable Jan 29 15:07:15 so we are trying to make changes a lot more visable Jan 29 15:07:27 if you want to, please add ideas to the wiki or the mailing lists Jan 29 15:07:46 one issue that has gotten a lot of attention as of late is the UI change to anaconda's UI Jan 29 15:07:50 * badkittydaddy (95a9941d@gateway/web/freenode/ip.149.169.148.29) has joined #fudcon-room-3 Jan 29 15:08:16 we have been working with the design team to redesign the interface Jan 29 15:08:22 * having issues with network Jan 29 15:08:32 * showing some of the discussions on the wiki (not sure of the URL) Jan 29 15:09:09 * redesign mockups on screen Jan 29 15:09:38 we've been trying to consolidate the pages of the installer and make it more modern-looking Jan 29 15:09:54 other distros have been doing better at this, and we want to improve Jan 29 15:10:13 * looking at wiki/anaconda/features Jan 29 15:10:30 * a list of all the features that are hoped for in anaconda Jan 29 15:10:46 it is a recent braindump, most of the recent ideas have been about testing Jan 29 15:11:05 some information at the earlier AutoQA meeting Jan 29 15:11:17 one of the most complicated parts of testing anaconda is storage testing Jan 29 15:11:39 it is very human-intensive to do the testing and even if you automated it, a human still has to check the system when its done Jan 29 15:12:11 its also is really hard to reproduce when you're doing installs and after 9k installs, you can't be sure of the initial state Jan 29 15:12:44 so the new method uses some kickstart and some virtualization (qemu) so that you can reproduce storage data in a generic sens Jan 29 15:13:12 everything is tied in through AutoQA, so that you could run a whole bunch of tests at every anaconda build Jan 29 15:13:37 another thing that is cool about the storage testing is that it makes things much easier to reproduce a env from user description Jan 29 15:14:23 so that we can test with almost the exact config that the user had (example - debian, ubuntu ... ,fedora; 12 different partitions - all sorts of obscure stuff) Jan 29 15:14:25 * Nushio has quit (Quit: Page closed) Jan 29 15:14:48 the anaconda storage stuff isn't on the wiki yet, all in chris Lumen's head Jan 29 15:14:55 * bug him to get stuff on the wiki :) Jan 29 15:15:23 some of the anaconda storage testing stuff is complicated, so it could be difficult to just jump into Jan 29 15:15:41 but hopefully we can advance the state of anaconda testing and make things easier to debug Jan 29 15:15:57 now we're going to start calling out individual contribs on the anaconda team Jan 29 15:16:04 Will Woods is up first Jan 29 15:16:12 * misha37 (95a9862d@gateway/web/freenode/ip.149.169.134.45) has joined #fudcon-room-3 Jan 29 15:16:21 in brief: pygtk is going to go away eventually Jan 29 15:16:34 and will be replaced with a new method of creating bindings with GTK Jan 29 15:16:46 *q - so is there going to be a new way of getting the bindings Jan 29 15:16:59 I want to say that they are similar, but not exactly Jan 29 15:17:16 so it should be a simple change for most, but there might be some problems Jan 29 15:17:34 and since we need to go through the code again, its a good time to revisit design decisions Jan 29 15:18:08 and its a good time to do this, since we are re-designing its good to do both, too Jan 29 15:18:39 So we're going to have a visual improvement with anaconda, and get rid of some of the ugly UI code at the same time Jan 29 15:18:55 and glade-up a bunch more things Jan 29 15:19:08 *q - does that give us the GSAPI stuff with glade Jan 29 15:19:10 in theory, yes Jan 29 15:19:37 the magic of gobject introspection, you magically get all the features that are in GTK Jan 29 15:19:51 *q - its just putting another layer between us in python and GTK? Jan 29 15:19:54 yeah, pretty much Jan 29 15:20:18 so, there is another big problem in anaconda Jan 29 15:20:45 anyone who has tried to generate a tree install (missed the rest of the sentance) Jan 29 15:20:59 tree composition for anaconda Jan 29 15:21:17 there is some new work to be able to generate a tree composition outside of a tree Jan 29 15:21:21 using a project called lorax Jan 29 15:21:31 it is a transparant change, but big for us Jan 29 15:22:01 the build install scripts included a root filesystem creation script Jan 29 15:22:16 there was some ugly lists of env vars for all of the tests Jan 29 15:22:30 it was a static list of file names and a static list of package names Jan 29 15:22:41 if anything changed, it won't fail but the images won't work Jan 29 15:23:09 *q - what we're talking about is a method of creating the OS subdir on the download site Jan 29 15:23:11 ? Jan 29 15:23:23 actually, it is the thing that becomes the installation CD Jan 29 15:23:46 and something more (missed what was actually said) Jan 29 15:23:55 *q - what happendd to the list of pkgs and files Jan 29 15:24:05 it has been translated into a list that lorax reads Jan 29 15:24:22 the problem was that yum wasn't used for pkg installation (in the old method) Jan 29 15:24:31 lorax does use yum, which makes everything easier Jan 29 15:24:38 you can actually tell waht is going on Jan 29 15:25:05 *q - there was a F14 bug taht broke the installer for a while; there was something from python .6 to 2.7 Jan 29 15:25:18 there was a magic file taht was needed and changed - which broke the installer Jan 29 15:25:34 does that mean that the installer is RPM based now instead of the old file list method? Jan 29 15:25:49 getting there - but now its whitelist based instead of just list based Jan 29 15:25:59 its not something taht has been specifially addressed ... Jan 29 15:26:21 *q - do i need to more fine-grain my pkgs to decrease bloat Jan 29 15:26:27 yes, that would be nice Jan 29 15:26:35 * - who do I talk to about this? Jan 29 15:27:04 its a tricky question Jan 29 15:27:17 building the installer images is a tricky issue (corner case?) Jan 29 15:27:27 ==> will get more into this issue during hackfest Jan 29 15:27:51 *q _ going from script to lorax, has the image gone from 20MB to 300mb? Jan 29 15:27:59 that is a change, removed the stage2 img Jan 29 15:28:09 *q - can we make this smaller? Jan 29 15:28:26 before, the initrd was 20mb, image was 150mb Jan 29 15:28:52 - mistake, total 120MB Jan 29 15:28:58 now lorax makes ` 130MB Jan 29 15:29:12 not `, ~ Jan 29 15:29:43 so combined, the total used space is decreased even though the individual files increased in size Jan 29 15:29:55 like not shipping the 150MB locale db Jan 29 15:30:08 -> want to talk about the network changes in anaconda Jan 29 15:30:17 anaconda uses networkmanager for install now Jan 29 15:30:26 * talk about old stuff for network in anaconda Jan 29 15:30:43 NM allows network functionality in anaconda without much extra work Jan 29 15:31:10 NM is only in stage2, not in stage1. stage1 only has basic support Jan 29 15:31:32 since stage1 is so small, we have to communicate with NM behind the scene to produce our own interfaces Jan 29 15:31:39 wireless can also be a challenge Jan 29 15:31:48 * there will be more discussion at the hackfest Jan 29 15:32:13 with NM, we no longer have the forced "configure your network" screen Jan 29 15:32:23 (this seems to have been controvertial) Jan 29 15:32:46 by default, we have install network config based on what happened during install Jan 29 15:33:01 this can be a problem when doing a media install Jan 29 15:33:18 *q - why isn't the nework config on by default Jan 29 15:33:29 some people didn't want to enable all interfaces on a server at install time Jan 29 15:33:50 -- missed the question, something about smolt? Jan 29 15:34:15 *q - so people who are installing to get a desktop at the end, having a disabled network is BAD Jan 29 15:34:24 * badkittydaddy has quit (Ping timeout: 265 seconds) Jan 29 15:34:34 *qcontd - would people doing a server install really use the GUI for install Jan 29 15:35:03 we recently stripped down the text mode, started recommending VNC re-install Jan 29 15:35:34 with the UI redesign, we have two paths for install - server (with lots of control) and a desktop (next, next, next) Jan 29 15:36:18 *comment - i think that one of the problems with the storage change is that not everyone knows the difference between basic and advanced storage Jan 29 15:36:43 *q - shouldn't the default desktop install download updates from yum? Jan 29 15:36:56 no, there are also people who do net installs that aren't on the internet Jan 29 15:37:22 *q - do people actually use the add repo stuff? Jan 29 15:37:24 yes, there are Jan 29 15:37:37 cause we want to take them away (Pain to maintain, test) Jan 29 15:37:56 ==> not actually, taking them away, was joke ( I think) Jan 29 15:38:16 *c - I think taht one of the issue with using NM in anaconda is that it can be akward Jan 29 15:38:47 some other people might be able to benefit from UI redesign (system-config-something for example) Jan 29 15:38:58 since we just use their UI, we can't really change it Jan 29 15:39:10 *c - NM isn't terribly easy to use other than through GUI Jan 29 15:39:27 supposedly there is a new NM CLI but haven't looked at it yet Jan 29 15:39:45 *q - ipv6 support in anaconda ui? Jan 29 15:39:49 everyone seems to think so Jan 29 15:40:10 dave lehman was primarially responsible for the storage refactoring Jan 29 15:40:23 he has been working on some new things, testing framework etc. Jan 29 15:40:59 the only notworthy thing that he can think of is the ability to install to a disk image file (qemu disk image) from anaconda Jan 29 15:41:37 so if you want to do a F14 install, "yum install anaconda" and do some other stuff (missed it ) and end up with a image on your drive with a new F14 image Jan 29 15:41:45 *q - how does that work with HD detection Jan 29 15:41:54 there is no HD detection Jan 29 15:42:08 *q - how would you then boot the image in KVM after you're done Jan 29 15:42:22 frankly, I don't know much about KVM, never even written the XML files that you would need Jan 29 15:42:35 in anaconda, we just have the basics, grub, disk etc. Jan 29 15:42:44 we don't care a whole lot about the details of the HW Jan 29 15:42:56 the kernel should deal with the HW details at runtime Jan 29 15:43:10 every piece of HW supported by fedora should be in the initrf Jan 29 15:43:12 initrd Jan 29 15:43:28 just point KVM at the image and boot Jan 29 15:43:44 we're also looking at some appliance creation stuff Jan 29 15:44:20 the live images have been very popular, want to make the livecd installation method as good as possible Jan 29 15:44:30 that method is a little bit of a hack at the moment Jan 29 15:44:39 ==> taking about livecd tools Jan 29 15:45:01 the current goals are to start using standard tools (ie anaconda) Jan 29 15:45:23 right now we have different install paths and we want to consolidate into a single path so that bugs only have to be fixed in one place Jan 29 15:45:38 which should lead to fewer installer issues and fewer bugs Jan 29 15:46:01 basically does a dd to copy an image to the system and expands the fs Jan 29 15:46:18 *c - you might not be able to get away from that -- very fast which is what people want Jan 29 15:46:36 *c - there are actually 2 dds Jan 29 15:47:06 *c - tried to teach middle school girls, 4 of 12 had catastrophic failure installing livecd Jan 29 15:47:09 really? Jan 29 15:47:40 they had persistant home directories on a liveusb Jan 29 15:47:52 it took 2 weeks to build these liveusb keys Jan 29 15:47:59 were you using the windows tool? Jan 29 15:48:05 no - never been able to get that to work Jan 29 15:48:18 -> note, this is a conversation, don't know names Jan 29 15:48:45 an idea is to gen sha1 sums at usb key egeneration time, check the sum after making a bunch of copies Jan 29 15:49:15 there are a bunch of layers where things can go wrong Jan 29 15:49:40 I feel bad that the live media is being touted so much but there are so many problems - potentially leading people into a false sense of security Jan 29 15:49:47 ==> back to presentation Jan 29 15:50:02 in the last couple of minutes, want to talk about a message to email list Jan 29 15:50:10 want to have anaconda around in multiple thread Jan 29 15:50:17 * audience "oh no" Jan 29 15:50:40 using GTK UI stuff, they tend to do a lot fo wor in the GTK codecs Jan 29 15:50:51 lot of work Jan 29 15:51:07 The design is that GTK can only work if its main loop is running Jan 29 15:51:25 to make things work in one thread, there is a hack where "GTK, you can do two loops now" Jan 29 15:52:00 and if the user gets bored looking at the progress bar, move the window around (live install, I think) Jan 29 15:52:22 this can cause problems because of the thread issues Jan 29 15:52:35 it would cause a lot of problems, but the idea would be that it would make the UI much more responsive Jan 29 15:53:11 example - the screensaver came on during install, screen was grey until next package triggered repaint Jan 29 15:53:27 * threethirty (~wircer@173.152.29.64) has joined #fudcon-room-3 Jan 29 15:53:31 it sounds like we're going to be doing an entire UI rewrite by F17 Jan 29 15:53:36 and we might Jan 29 15:53:51 if you're interested, join in the conversation on the wiki Jan 29 15:54:01 fedoraproject.org/wiki/Anaconda/Feature Jan 29 15:54:16 tomorrow, during the hackfest sessions, we'll be discussing a lot of these features Jan 29 15:54:28 if you're interested, let us know and we'll include you on the conversation Jan 29 15:54:53 so that's some of the major stuff going on in anaconda right now, condensed into an hour Jan 29 15:55:08 so anyways, thank you for coming Jan 29 15:55:11 * misha37 has quit (Ping timeout: 265 seconds) Jan 29 15:55:18 ** applause and group photo for anaconda folks Jan 29 15:55:26 ====>> end of anaconda presentation Jan 29 15:55:57 I will end up putting my logs of the presentation up on the wiki later today/tonight