2010-09-06 01:41 xiangfu: hi! how is your success with the latest build? 2010-09-06 01:41 mine is not very good 2010-09-06 01:41 NanoMap is not working (complains that now driver could be found, lookinf for /usr/bin/gfxdrivers/) 2010-09-06 01:41 startdict is hanging is before, during startup 2010-09-06 01:42 nlove couldn't be started, too 2010-09-06 01:44 oh 2010-09-06 01:44 that's not a problem with the image though 2010-09-06 01:44 the default screen you see when there are no arguments passed is still 800x600 2010-09-06 01:44 bartbes: yeah.. it complains something about booter.lua 2010-09-06 01:44 so it can't open the display 2010-09-06 01:44 or.. that 2010-09-06 01:44 if you feed it a game it'll work 2010-09-06 01:44 hm, how do i do it? 2010-09-06 01:44 i just 'nlove' 2010-09-06 01:45 kyak: :( mine is still compiling.Â Â I start is 4 hours before. 2010-09-06 01:45 well, first you get a game, so npong: http://dl.dropbox.com/u/440010/nlove/npong.love 2010-09-06 01:45 or snake: http://dl.dropbox.com/u/440010/nlove/snake.love 2010-09-06 01:45 then you run nlove 2010-09-06 01:46 I should fix the no game stuff sometime 2010-09-06 01:46 :P 2010-09-06 01:49 bartbes: ok, thanks! gonna give it a try :) 2010-09-06 01:49 for some reson, i thought that "nlove" IS the game :) though i remember you explained it's just a framework 2010-09-06 01:51 xiangfu: well, ok, maybe you'll have a better luck :) 2010-09-06 01:52 xiangfu: is your latest config commited to git? 2010-09-06 01:52 kyak: no. since it's still compiling. 2010-09-06 01:53 i mean, the one from 3rd September 2010-09-06 01:53 or when you made the previous build 2010-09-06 02:22 kyak: hmm.. I am not keep that one. 2010-09-06 02:23 xiangfu: I think you should commit the config file before you even start the build. what is the advantage of keeping it local? 2010-09-06 02:23 the revision control system can handle a few revisions :-) 2010-09-06 02:24 wolfspraul: I want test before commit. :) 2010-09-06 02:24 wolfspraul: I don't think without test , then just commit is a good idea. 2010-09-06 02:25 yes but it's a config file, and it needs to be copied to .config manually 2010-09-06 02:25 so whatever you commit, even if you commit a 'broken' config file, the build process that involves it is still manual 2010-09-06 02:25 wolfspraul: ok. 2010-09-06 02:25 it's more like a documentation file, give its position in the directory tree 2010-09-06 02:26 well that's my opinion only, if someone disagrees fine. do what you think is right. 2010-09-06 02:26 wolfspraul: understand. :)Â Â thanks. 2010-09-06 02:26 but you see there were several people already asking now, and you send the file by mail, or you don't have it anymore, etc. 2010-09-06 02:26 sounds like if you would use revision control, you could point everybody to what they are asking for 2010-09-06 02:27 wolfspraul: yes. 2010-09-06 02:27 and you don't break the build, because this file is in a manual-use location in the tree anyway 2010-09-06 02:27 wolfspraul: yes 2010-09-06 02:27 wolfspraul: I have add a sample music file to /root/Music 2010-09-06 02:28 [commit] Xiangfu Liu: add some sample music file. http://qi-hw.com/p/openwrt-xburst/e39e43e 2010-09-06 02:28 [commit] Xiangfu Liu: update config to use uClibc 0.9.30.1, change VERSION to 2010-09-06 http://qi-hw.com/p/openwrt-xburst/0b9c16f 2010-09-06 02:28 ok 2010-09-06 02:28 we need to think about that 'including data' problem later in more detail 2010-09-06 02:28 but some sample files are good for now 2010-09-06 02:29 wolfspraul: I follow the roh's suggest. put a README file and attach the sample URL. license. 2010-09-06 02:29 later if we want to include more data, say entire dictionaries, wiki content, maps, etc. we need a way to package them. should we use normal openwrt .ipks for the data? or some other mechanism? I don't know. 2010-09-06 02:29 sure that's OK for now, that's about satisfying the attribution requirement. 2010-09-06 02:30 I am talking about larger amounts of data now, say entire offline wikis, or larger amounts of map data. 2010-09-06 02:30 how should they come to the nano, packaged in which way? openwrt is typically only used for binaries (executables), maybe some small amount of documentation files 2010-09-06 02:31 but entire music albums? hundreds of megabytes of data? like I said don't know. 2010-09-06 02:32 wolfspraul: hmm.. me either. 2010-09-06 02:34 slackware tgzs 2010-09-06 02:37 bartbes: slackware tgzs what? can you be more precise? 2010-09-06 02:37 I think the normal packaging systems are suited for binaries and especially dependencies/versioning of packages. 2010-09-06 02:37 it's just a tgz but it includes full paths 2010-09-06 02:37 so if you want to install something to /usr/share/gmenu2x/epic-wallpaper.png 2010-09-06 02:38 so what? we put such files on the server somewhere? how do they get installed, how uninstalled? are updates necessary? 2010-09-06 02:38 for example offline wiki content, dictionaries, map data 2010-09-06 02:38 you have a tgz containing a folder usr, which contains share, which contains gmenu2x which then contains the wallpaper 2010-09-06 02:38 cannot have all maps of the whole world on the nanonote 2010-09-06 02:38 to install you just extract 2010-09-06 02:38 say for nanomap 2010-09-06 02:38 of course removing is.. somewhat.. harder 2010-09-06 02:38 let's say you go to a trip in nothern italy 2010-09-06 02:39 florence, venice, verona 2010-09-06 02:39 how can you quickly get all of northern italy onto your nanonote? 2010-09-06 02:39 and when you are back home, do you quickly delete it 2010-09-06 02:39 yes, I know what you want to do 2010-09-06 02:39 or say offline english wiki data 2010-09-06 02:39 here the main issue is probably updates 2010-09-06 02:39 how to update it? 2010-09-06 02:39 sure we can throw everything into .ipk and reuse that system 2010-09-06 02:40 a super-intelligent bash (ehm.. ash) script 2010-09-06 02:40 but : 1. that's definitely a different usage of the packaging system than what anybody upstream would use it for. 2. it may not be suited well to the needs of large amounts of data, incremental updates, etc. 2010-09-06 02:41 for example I would like to quickly have a couple of Chinese cities in nanomap 2010-09-06 02:42 the whole guangzhou/shenzhen/hongkong section :-) 2010-09-06 02:42 well we see, don't need to find a perfect system today 2010-09-06 02:42 I'd suggest a vcs if they weren't so inefficient (with space) 2010-09-06 02:43 but as the software gets better (the viewers & editors), the question of how to get larger amounts of data to the ben and removed becomes more important 2010-09-06 02:43 otherwise why have viewers? 2010-09-06 02:43 :-) 2010-09-06 02:43 for larger amounts, incremental updates would also be cool 2010-09-06 02:43 squashfs 2010-09-06 02:44 ofc you would have to mount a lot 2010-09-06 02:44 my feeling is data is just different from binaries, maybe some solutions already exist 2010-09-06 02:44 but you'd get self-contained data 'files' 2010-09-06 02:44 the normal packaging systems are geared towards the needs of binaries 2010-09-06 02:44 heavy on dependencies, versioning 2010-09-06 02:45 bartbes: what do you mean with vcs? 2010-09-06 02:46 sorry that abbreviation doesn't ring a bell 2010-09-06 02:46 version control system 2010-09-06 02:46 (git, hg, etc) 2010-09-06 02:46 most are really weak with binary data 2010-09-06 02:46 they are inefficient anyway 2010-09-06 02:46 but in a lot of cases we may be dealing with binary data 2010-09-06 02:47 so, I then said squashfs 2010-09-06 02:49 I don't know. how does that help with the delivery problem, incremental updates, removal of data. 2010-09-06 02:49 well, removal is easy 2010-09-06 02:49 just remove the sfs file, done 2010-09-06 02:49 updates are a problem, yes 2010-09-06 02:49 I think we need some software, just like dpkg/opkg, but tailored to the needs of data 2010-09-06 02:49 but at least it contains all data in a 'package' 2010-09-06 02:50 I don't believe in twisting tools that were made for a particular purpose to do something else. 2010-09-06 02:50 I cannot believe that we are the first ones to look into this problem. 2010-09-06 02:50 the nanonote has 2gb today, not that much 2010-09-06 02:50 and tomorrow? 16gb, 32gb, etc? 2010-09-06 02:50 what to do with all this space? 2010-09-06 02:50 nothing? 2010-09-06 02:51 just need some good 'data manager' that knows what type of data it is managing, can get updates, can visualize amounts of data, make it easy to remove etc. 2010-09-06 02:51 what data are we talking about? 2010-09-06 02:51 music, videos, dictionaries/wiki content, maps 2010-09-06 02:51 more? 2010-09-06 02:53 games :P 2010-09-06 02:53 books 2010-09-06 02:53 pictures 2010-09-06 02:53 yes sure, games. but many of those are still well suited to be managed by normal package managers, no? 2010-09-06 02:53 (/images) 2010-09-06 02:53 it seems a bit useless to make a ipk for a single .love file 2010-09-06 02:54 sure but then you just package suites of games 2010-09-06 02:54 unless the data content becomes really large, I think normal .ipks will just do fine 2010-09-06 02:54 what I am talking about is more about music albums, maps, etc. 2010-09-06 02:54 for a music album, there will never be dependencies 2010-09-06 02:55 nor will there be an 'update' or a 'new version' 2010-09-06 02:55 true 2010-09-06 02:55 there may be different music. I may get bored with some album. 2010-09-06 02:55 but for maps there will be updates 2010-09-06 02:55 yes 2010-09-06 02:55 but maps will be so big, you cannot have the whole world on your nano 2010-09-06 02:55 to it should be easy to swap in and out, for trips 2010-09-06 02:56 and it should be easy to update, if you are in a region a bit longer, say your home region 2010-09-06 02:56 but what I meant is, that you should not rule out versioning 2010-09-06 02:56 also let's say your device is full, but you need some free space, the data manager should be able to quickly help you free a couple hundred meg 2010-09-06 02:57 also, I wondered.. wouldn't it be great if it were capable of installing straight from your computer? 2010-09-06 02:57 saying "there is these 5 music albums that you haven't listened to in a year anyway" - click here to delete 2010-09-06 02:58 the data manager should present me with 'unused data', the same way dpkg can tell me about orphaned or unused packages 2010-09-06 02:59 big diff 2010-09-06 02:59 it does that based on deps 2010-09-06 02:59 bartbes: yes, fine [versioning] but my point is, if the people writing the tool, say opkg, don't really wrap their head around the needs of data, in the use cases we are talking about here, then abusing their tool for data may work for a while, but will never be really god 2010-09-06 02:59 not on whether you actually use it or not 2010-09-06 02:59 anyway 2010-09-06 02:59 gt 2010-09-06 02:59 *gtg 2010-09-06 02:59 and if you explain your problems to them, they will be like "oh, we never use our tool like this" 2010-09-06 03:00 I don't feel right packaging a 100 mb music album into an .ipk file 2010-09-06 03:00 does anybody else do that? 2010-09-06 03:03 there's some activity in distributed filesystems, but not sure whether it applies 2010-09-06 03:03 I don't believe massaging this into sql databases either, they were built for relational data... doesn't fit. 2010-09-06 03:07 are any of the distributed file systems any good nowadays? andrew, coda, etc? do they apply to this problem? 2010-09-06 04:27 Ornotermes: hmm, have you tried drawing power from VDD ? I'm getting ridiculously little current 2010-09-06 04:30 wpwrak: i got a slow voltage drop when i connected my scope, so i suspect the fet transistor is switched off 2010-09-06 04:30 wolfspraul: the real solution for xiangfu's slow updates will be to get him a faster PC :) also, even if the config is copied manually, if I understand what this is about, people who follow the manual process would still pick up the untested file, no ? 2010-09-06 04:31 wpwrak: check the state of PD02 2010-09-06 04:32 I see the voltage rise and fall like it should. but when i add any kind of significant load, it drops badly. 2010-09-06 04:38 i modified an otherwide empty board to have a 100 Ohm resistor between VDD and GND. voltage drops to about 66 mV. that can't be right. 2010-09-06 04:39 oh, wait. i have GPD06, not GPD02. hmm :) 2010-09-06 04:40 i traced it to PD02 anyway, but i could ofc be wrong :P 2010-09-06 04:42 PD02 definitely looks right. 2010-09-06 04:42 not sure where i got the GPD6 from 2010-09-06 04:57 wpwrak: if we only commit the config file after a successful image release, we have created a huge bottleneck to ever improve the image in the sense that more stuff is included 2010-09-06 04:58 in fact, this is why after over a year, we still don't have at least 50 easily included apps included, and sometimes even some apps 'drop out' 2010-09-06 04:58 it must be the other way round. whenever someone commits an app that compiles, the respective line in the config file is set to Y, and the config file is committed 2010-09-06 04:58 if someone builds an image later, and some apps don't compile, it's easy to set them to N again 2010-09-06 04:59 the other way round simply means that there will never be progress in completing the apps that are included 2010-09-06 04:59 or maybe it takes a few centuries, at the speed we give feedback such as 'sox is missing', and then it takes x months until that is reflected in the image 2010-09-06 04:59 so like I said - everybody should feel free to 'enable' a nice little app in the config file, and commit that 1-line change right away 2010-09-06 05:00 at least until we feel we have a really strong set of 'base apps' in the image and running out of space in the rootfs 2010-09-06 05:03 will people who follow the build instructions pick up such untested configs ? if yes, that would be bad 2010-09-06 05:03 currently we have 2 config files (remember, at that location the configs are more like documentation / starting points) 2010-09-06 05:03 one "stable" and one "testing" ? 2010-09-06 05:04 one is for xbboot (rescue image to be booted via usbboot), and the other one is an image that should include as complete as possible set of 'base apps' 2010-09-06 05:04 we could have a third config file, empty base system 2010-09-06 05:04 config.base 2010-09-06 05:04 config.factory_default 2010-09-06 05:04 config.xbboot 2010-09-06 05:05 but when would someone need config.base? I don't know 2010-09-06 05:05 seems that you need a stage for yet to be tested configs. that way, people who can't wait for xiangfu's pc can pick the bleeding edge config, while those who just want to build anything recent will get the stable one 2010-09-06 05:05 no no. I don't see the config files as having anything to do with development. it's not about unstable/testing/stable 2010-09-06 05:06 it's just a list of apps, Y or N 2010-09-06 05:06 so the xbboot image needs to be small 2010-09-06 05:06 a few megs 2010-09-06 05:06 and only needs tools you would want in a rescue situation 2010-09-06 05:06 since this is an initramfs that is loaded to boot/unbrick a nanonote directly from usbboot 2010-09-06 05:06 the smaller the better (in that case) 2010-09-06 05:07 the factory default should make some use of the NAND, which otherwise is just empty, and include all meaningful apps contributed or tailored for the Ben NanoNote 2010-09-06 05:07 Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â reflected in the image 2010-09-06 05:07 (sorry) 2010-09-06 05:07 did a config change ever have to be reverted due to problems ? 2010-09-06 05:07 of course, whenever an app actually doesnt' compile 2010-09-06 05:07 aha ! so it is "development" :) 2010-09-06 05:07 most of the time, when someone wants to make an image, and app whatever doesn't compile, that person will just kick that app out of the list 2010-09-06 05:08 it's an app list 2010-09-06 05:08 there is no concept of unstable/testing/stable here 2010-09-06 05:08 but that reference list should be reliable, no ? 2010-09-06 05:08 if a kernel module doesn't compile, people will just disable it and build the rest 2010-09-06 05:08 there are many kernel modules :-) 2010-09-06 05:08 unless that is a truly essential one that won't even boot the system 2010-09-06 05:08 i mean, if i just want to build the whole stuff, i don't really want to have to debug the thing too 2010-09-06 05:09 people deal with this config file the same way 2010-09-06 05:09 of course if uclibc doesn't build, the image build cannot proceed 2010-09-06 05:09 but if yacas doesn't built, it's just set to 'N' for this time 2010-09-06 05:09 what is 'whole stuff'? 2010-09-06 05:09 sure, people can work around this. but why is it a benefit if they have to ? 2010-09-06 05:10 well, build the "standard system" 2010-09-06 05:10 we may need different config files 2010-09-06 05:10 ok, I described the xbboot system to you 2010-09-06 05:10 :-) 2010-09-06 05:10 that one is easy 2010-09-06 05:10 the smaller the better 2010-09-06 05:10 rescue tools 2010-09-06 05:10 then we need a factory image 2010-09-06 05:10 xboot config may not change all that often anyway 2010-09-06 05:10 that one should include all meaningful ben nanonote apps 2010-09-06 05:11 yup, that one 2010-09-06 05:11 don't you see that if you don't commit first, and then remove apps that don't build, the config file will simply never get updated? 2010-09-06 05:11 because, big surprise, it will always work, nothing will ever break 2010-09-06 05:11 only that all apps are missing :-) 2010-09-06 05:11 what's wrong with updating it after testing ? 2010-09-06 05:11 how do you want to break out of this circle? 2010-09-06 05:11 yes, but people work on apps one by one 2010-09-06 05:12 the problem is that they never get turned on in the config file 2010-09-06 05:12 make a "testing" variant and a "tested" variant 2010-09-06 05:13 those who just want a build take the tested variant. those who don't mind debugging the latest config take the untested one. xiangfu seems to do the full build regularly anyway, so there is a process for pushing the untested to the tested version. 2010-09-06 05:14 are you saying that people working on apps don't cross-compile them themselves with "make" but use openwrt as their IDE ? 2010-09-06 05:14 please, not that insanity again :) 2010-09-06 05:14 I think this idea will successfully reduce the commits into both variants, because now people will be unsure which one to commit to. 2010-09-06 05:14 the config file is just an app list 2010-09-06 05:15 a makefile is also a list of object files :) 2010-09-06 05:15 whenever someone works on an app, and gets that app to work, the last step is to enable the app in the config file 2010-09-06 05:15 when someone builds an image later, and out of 20 apps 17 build, and 3 break, then those 3 need to be removed (set to 'N') at that point. 2010-09-06 05:15 and it doesn't matter what the config file is. the issue is that it can break. 2010-09-06 05:16 so you say there's no build process that can be trusted to work, and that's even intentional ? 2010-09-06 05:17 the other way round. when someone completes work on an app, they know it builds. why would they not set the app to 'Y' in the app list now? 2010-09-06 05:17 because they don't trust the build process in general? 2010-09-06 05:17 [commit] Xiangfu Liu: [poke] remove BROKEN, update to r5975 http://qi-hw.com/p/openwrt-packages/807277d 2010-09-06 05:17 and even though it works on their own system now, they just keep it disabled? 2010-09-06 05:17 the use scenario i have in mind is someone who just wants to have a reasonably up to date system 2010-09-06 05:18 they probably disable the rest locally to get a faster build 2010-09-06 05:18 wpwrak: ok good, which use case do you have in mind? 2010-09-06 05:18 someone wants to make a new image, why? 2010-09-06 05:18 I have two in mind right now, xbboot and factory default. 2010-09-06 05:18 to get new things. features and fixes. update the system from time to time. 2010-09-06 05:18 why wouldn't they just reflash from the published images? 2010-09-06 05:19 (actually I hope we can move to upgrades via packages, but that's another story...) 2010-09-06 05:19 we are talking about a config file here that defines which apps are included when that config file is used 2010-09-06 05:19 an app list 2010-09-06 05:19 we can have multiple app lists 2010-09-06 05:19 (from images) well, there's admittedly that option. makes them quite blind, though. 2010-09-06 05:19 yes sure, so they want to compile from source 2010-09-06 05:19 but why? 2010-09-06 05:20 if they want to compile the exact same image they had on their device, they can check /etc/VERSION and get the matching config file which is always kept right next to the image file anyway 2010-09-06 05:20 totally outside the revision control system 2010-09-06 05:20 source build is also so far the only way to get a toolchain 2010-09-06 05:20 so we are not talking about that case 2010-09-06 05:20 true, we need to look into that (toolchain), but what does that have to do with the config file? 2010-09-06 05:21 (matching config file) ah, okay. so there _is_ a place for the "tested" config 2010-09-06 05:21 it's not a default config file, it's just a config file at a place in the tree where it needs to be (can be) manually taken from 2010-09-06 05:21 yes 2010-09-06 05:21 (toolchain) the config file is still necessary for the openwrt build, no ? 2010-09-06 05:21 at least that's how i understood the build process 2010-09-06 05:22 yes it selects the target too 2010-09-06 05:22 one sec look here 2010-09-06 05:22 http://downloads.qi-hardware.com/software/images/NanoNote/Ben/2010-06-15/ 2010-09-06 05:22 that's a released image 2010-09-06 05:22 the config file that was used to create this image is right there with it, together with all other things that hopefully make it possible to recreate this exact image 2010-09-06 05:23 alright. that one has a config. good :) 2010-09-06 05:23 (although I have never tried and I thnk it's hard with all the many versions updates everywhere) 2010-09-06 05:23 so in a way, that's the 'tested' config you talked about before 2010-09-06 05:23 you are right, thinking about toolchain 2010-09-06 05:23 but it's painful that someone has to build all factory default apps only to get the toolchain 2010-09-06 05:23 okay, looks good then 2010-09-06 05:23 (toolchain) indeed 2010-09-06 05:24 we could have a config file config.base 2010-09-06 05:24 or config.toolchain 2010-09-06 05:24 and even worse if they have to fix any config glitches :) 2010-09-06 05:24 of course, total nonsense 2010-09-06 05:24 how about making a toolchain tarball, like we had at openmoko ? i think that one worked rather nicely 2010-09-06 05:24 but I am talking about the app list in the factory images, and our very painful process of ever getting that app list to grow to include all useful apps 2010-09-06 05:24 which at the current speed will not happen for decades :-) 2010-09-06 05:25 (app list) is it really that bad ? isn't it just those ~10 hours it takes xiangfu to build the image ? 2010-09-06 05:26 yes but we are committing config file updates too slowly (this is how this discussion started) 2010-09-06 05:26 I think apps must be enabled in the config file at the same time the (working) app Makefile is commited 2010-09-06 05:26 otherwise why work on the app? and who will later remember to enable that app? 2010-09-06 05:26 it can't work, already proven for x months now 2010-09-06 05:27 xiangfu actually has a quite powerful machine, if it's only 10 hr for the whole thing that's quite good 2010-09-06 05:27 so what's the solution now? we have realized that there are in fact 'stable/tested' config files, together with the released images 2010-09-06 05:27 (who gets to enable the app) ok, i see what you mean. 2010-09-06 05:28 we may rename that 'config' file to 'config.factory_default'? 2010-09-06 05:28 and change the commit policy into that config file to 'first set to Y when app is committed, then remove when it breaks'? 2010-09-06 05:29 create another config file for a base/empty system, maybe for people who only want the toolchain? 2010-09-06 05:29 and the build instructions have a "stable" config, correct ? (http://en.qi-hardware.com/wiki/Building_Software_Image) 2010-09-06 05:29 (there may be better ways to do this in OpenWrt, have to check) 2010-09-06 05:30 i have added CC BY-SA for all my nanonote related pictures now: http://gallery.slashhome.se/main.php?g2_itemId=2616 and http://gallery.slashhome.se/main.php?g2_itemId=2856 2010-09-06 05:31 wpwrak: yes it points to the stable/tested one 2010-09-06 05:31 which is correct 2010-09-06 05:34 Ornotermes: nice. your pictures are great, I really like them. 2010-09-06 05:34 wolfspraul: okay, perfect then 2010-09-06 05:35 wolfspraul: thanks! :) 2010-09-06 05:35 so the config in git is really meant to be "unstable at all times", but we have a nice and tested one as well. 2010-09-06 05:36 you can see it like that 2010-09-06 05:36 I think we should rename it to config.factory_default 2010-09-06 05:36 i was missing that last point. that works then. thanks for explaining :) 2010-09-06 05:36 ah thanks for pointing out the loose ends 2010-09-06 05:37 and encourage people to enable their apps in config.factory_default if they think it should be included in every NanoNote 2010-09-06 05:37 hmm, "factory default" sounds a bit like "fall back to minimum", which isn't the case here 2010-09-06 05:37 ok then what? 2010-09-06 05:37 config.factory_full 2010-09-06 05:37 :-) 2010-09-06 05:38 full_system ? just full ? 2010-09-06 05:38 "factory" may also scare people 2010-09-06 05:38 true, I thought about that too. so config.full_system ? 2010-09-06 05:38 sounds good to me 2010-09-06 05:38 alright, great 2010-09-06 05:39 let me see what the easiest way is to build just the toolchain 2010-09-06 05:39 (we can still offer a binary toolchain, that's separate) 2010-09-06 05:39 you don't like the tarball idea ? 2010-09-06 05:39 there are 2 options in OpenWrt 2010-09-06 05:39 ah. you anticipated the question :) 2010-09-06 05:40 "Build the OpenWrt SDK" 2010-09-06 05:40 "Build the OpenWrt based Toolchain" 2010-09-06 05:40 ironically, both seem to be OFF in our config files :-) 2010-09-06 05:40 maybe the toolchain is automatically built as a dependency? or the "OpenWrt based Toolchain" is something else? 2010-09-06 05:40 need to find out... 2010-09-06 05:51 sounds interesting :) the toolchain is built and stays around with the "standard" build, so that works 2010-09-06 05:52 which standard build? 2010-09-06 05:52 I just built the SDK, it created a 25 MB .tar.bz2 file 2010-09-06 06:02 hmm, the SDK just seems to be a tarball of part of the OpenWrt tree. I am wondering whether it is not even better to tarball the entire OpenWrt tree then, including everything downloaded and pre-built as in the full_system image? 2010-09-06 06:03 that will be bigger in MB, but maybe a faster starting point? 2010-09-06 06:03 we are really talking about speed to get started with the various options. build everything from source, toolchain. Or download a tarball with toolchain binaries, and maybe also full_system sources and binaries all included? 2010-09-06 06:04 that's a couple hundred MB download probably 2010-09-06 06:04 or direct people to always start with git, then compile everything from source? 2010-09-06 06:05 personally I'm not a big fan of pre-built binaries (although I run Debian, not Gentoo) 2010-09-06 06:05 wpwrak: what do you think? what should we offer? 2010-09-06 06:06 if you are going to make big files, use bittorrent as an alternative download source 2010-09-06 06:10 if I select that second option 'OpenWrt based Toolchain', it gives me a 15 MB tarball containing just the toolchain I think 2010-09-06 06:41 done duilding date command fail :p 2010-09-06 06:42 Btw i noticed NanoMap is working now, i'll try test it with my non.free gps later 2010-09-06 06:44 well i noticed quite the opposite, that NanoMap is no longer working :) 2010-09-06 06:50 ah? 2010-09-06 06:50 i dint test it in the last-lastest image from xiangu 2010-09-06 06:51 but the lsat that the flash script isntal me had nanomap, i jut said wow ! 2010-09-06 06:58 kristianpaul, seem interesting....does nanomap works with Xorg? 2010-09-06 07:03 GNUtoo|laptop: nope fb as i now 2010-09-06 07:03 ok 2010-09-06 07:03 as openwrt dont work with fb and X and same time 2010-09-06 07:03 well, it can work in X. too 2010-09-06 07:04 if we had X 2010-09-06 07:04 nice 2010-09-06 07:04 ok 2010-09-06 07:04 at least it works on my laptop just fine 2010-09-06 07:07 nice 2010-09-06 07:29 well my feeling is that the OpenWrt SDK is a better starting point for people who want to port an app, and not start with the entire openwrt tree and build from scratch 2010-09-06 07:30 better than the plain 'OpenWrt Toolchain' which is missing libraries, if I see this right 2010-09-06 07:30 so maybe we should publish this .tar.bz2 SDK tarball along with the images? 2010-09-06 08:26 [commit] Xiangfu Liu: rename the config to config.full_system http://qi-hw.com/p/openwrt-xburst/49ac818 2010-09-06 08:27 wolfspraul: toolchain plus the essential libraries sounds good, yes 2010-09-06 08:28 I think we will publish both first, then ask what people like, then remove what they don't want 2010-09-06 08:28 wolfspraul: the definiion of "essential" can differ, of course :) 2010-09-06 08:28 I want to publish as few options as possible, because the problem is that the quality of documentation may matter more than what binaries are available for download somewhere 2010-09-06 08:29 if the documentation quality is junk, people will get lost in all the lousy documented options 2010-09-06 08:29 so rather have fewer options that are better documented and linked to 2010-09-06 08:29 but let's start with both, then remove one after a few weeks 2010-09-06 08:30 yes, people who want to get started will prefer not to have to have too many choices 2010-09-06 08:59 Ornotermes: power to the board now seems to work, but when switching it on, there seems to be enough inrush current to crash the ben :-( 2010-09-06 08:59 Ornotermes: and there's an interaction with the kernel driver, too: when it detects the card, it tries to power it up, which then of course also kills the system 2010-09-06 09:00 Ornotermes: i don't think 33 mA or so should really be a problem. let's see what's really going on there ... 2010-09-06 09:25 wpwrak: how mych can the voltage regulator supply? 2010-09-06 09:52 what is included in a openwrt SDK? 2010-09-06 09:54 there is a total of three options 1) toolchain 2) sdk 3) image builder 2010-09-06 09:54 I think basically they include more and more of the entire tree 2010-09-06 09:54 toolchain is just the cross-compiler 2010-09-06 09:55 I think you can only compare simple binaries with that 2010-09-06 09:55 'sdk' includes a lot more, libraries from the device rootfs, package making tools? 2010-09-06 09:55 image builder contains even more, I think basically all binaries that appear in the image, so you can make a modified image without rebuilding anything that was already in the image before 2010-09-06 09:56 I think we will just publish all three options, and then find out what people use, and remove the unused ones 2010-09-06 09:56 you can always do the 'gentoo style' and get the source tree from git, and build it all from scratch 2010-09-06 09:56 just takes a bit longer, couple hours maybe 1 day, depending on your machine 2010-09-06 09:57 this is my understanding right now, if I'm wrong please let me know 2010-09-06 09:59 yeah also i must be skilled in openwrt packaging wich is not so hard but takes time 2010-09-06 10:00 s/I/you 2010-09-06 10:00 hope thereare package makign tools in that :) 2010-09-06 10:10 [commit] Xiangfu Liu: select ImageBuilder, Toolchain, SDK http://qi-hw.com/p/openwrt-xburst/4c4035c 2010-09-06 10:10 [commit] Xiangfu Liu: enable the IPV6 option http://qi-hw.com/p/openwrt-xburst/1566c3b 2010-09-06 10:10 [commit] Xiangfu Liu: include [sox] and [poke] http://qi-hw.com/p/openwrt-xburst/19ee485 2010-09-06 10:10 [commit] Xiangfu Liu: switch uClibc 0.9.30 to 0.9.32 http://qi-hw.com/p/openwrt-xburst/2af4183 2010-09-06 10:10 [commit] Xiangfu Liu: include qt4-* package http://qi-hw.com/p/openwrt-xburst/b9f0c00 2010-09-06 10:14 sox good :) 2010-09-06 10:15 This is getting better every day :) 2010-09-06 10:16 ipv6 2010-09-06 10:16 that is what I'm liking 2010-09-06 10:16 btw, about pke 2010-09-06 10:16 *poke 2010-09-06 10:16 I imagine it's like the c64 poke 2010-09-06 10:16 but.. are there any fixed mem addresses? 2010-09-06 10:16 wolfspraul: one missing SDK is i think a easy (not bloated) way of UI design for some usefull software like sox that lacks one 2010-09-06 10:17