**** BEGIN LOGGING AT Sun Jul 11 02:59:57 2010 Jul 11 03:09:51 I guess it isn't an industrial temp part :-) Jul 11 03:11:58 ~~~ Jul 11 03:20:00 is anyone here using agar? Jul 11 03:20:28 hborders: yes Jul 11 03:20:39 tman_mac: sweet. you are a lifesaver Jul 11 03:20:44 hehe ;) Jul 11 03:21:00 tman_mac: does it work in opengl mode? Jul 11 03:21:13 http://pastebin.com/LGQmsjBc Jul 11 03:21:19 This is the file I'm working with right now Jul 11 03:21:31 notice the comment about AG_VIDEO_OPENGL Jul 11 03:21:43 hborders: No, it doesn't support GLES currently. However I have been considering adding a GLES driver Jul 11 03:22:09 right now it just uses some OpenGL features that don't exist in GLES.. though I suspect it could be ported easily Jul 11 03:22:20 tman_mac: ok, cool. I wish I knew GLES so I could help Jul 11 03:22:46 second question: how are you starting agar? Can you post some examples of your inits? Jul 11 03:22:49 for now the sdl_fb driver is probably fine for most applications Jul 11 03:23:03 lemme see Jul 11 03:23:10 I'm getting flicker right now Jul 11 03:23:29 also, the hello.c app was leaving cheese when I would drag the window around Jul 11 03:23:49 I'm a total noob at agar. I'll learn the api once I'm sure I can display a clean window Jul 11 03:24:24 do you think that gles would improve performance/efficiency or not make much of a difference? Jul 11 03:24:44 the way I'm using agar is integrating it into an existing SDL application, so I grab SDL events and feed agar the ones I think it needs (I recommend this so you can trap SDL_ACTIVEEVENTs and handle your app being deactivated/reactivated) Jul 11 03:25:42 hborders: if you have somthing that's being updated constantly, it could help.. if it's mostly static, it probably won't matter.. Jul 11 03:26:24 tman_mac: I'm writing a twitter client. When scrolling, I'm guessing there would be a performance improvement Jul 11 03:26:35 yes I could see that Jul 11 03:26:41 tman_mac: Ok, so you wrote a custom event loop then Jul 11 03:27:08 tman_mac: what about initializing agar? that's my main concern. What flags did you set, and what flags won't work at all Jul 11 03:29:03 pasting some code Jul 11 03:29:47 thanks Jul 11 03:30:05 I'm happy to dump all of this onto the webosinternals wiki Jul 11 03:30:26 and clean it up a little into a hello agar for the pre Jul 11 03:31:14 :) Jul 11 03:31:16 http://webos-internals.pastebin.com/3z2vfAkU Jul 11 03:32:21 note that I create a surface in memory for Agar instead of directly using the screen.. but you could do that too if you used SDL_SetVideoMode (with w,h being 0,0) instead of CreateRGBSurface Jul 11 03:32:55 (I do that because I'm using GLES elsewhere in my app.. if you aren't don't worry about it :p) Jul 11 03:33:29 and the EventToAG() function.. a little ugly but necessary. Taken directly from agar source (the developer recommended I do this) Jul 11 03:38:23 I'm just reading through and trying to process this. it makes sense though Jul 11 03:39:41 yeah. questions? Jul 11 03:40:31 well first off, do you mind me shrinking this to its essence and putting up a wiki article about it? Jul 11 03:40:40 this could be a really great start for other people like me Jul 11 03:41:01 no not at all.. that's great. I was thinking of doing that if I ever got around to it ;) Jul 11 03:41:46 I recommend keeping in the SDL_ACTIVEEVENT stuff as an example, since many applications need to do that Jul 11 03:46:42 yeah, that was my plan Jul 11 03:47:07 that has to do with the app getting card-ified right? Jul 11 03:47:19 yeah Jul 11 03:47:44 and it has to re-draw itself on a SDL_VIDEOEXPOSE event Jul 11 03:50:50 I think other code just does an sdl display call right? Jul 11 03:51:04 I suppose there's some agar equivalent to tell agar to redraw itself Jul 11 03:53:13 hborders: yeah. for non-gl you'd probably use SDL_Flip(). for agar look for AG_WidgetUpdate() (I think) Jul 11 03:53:17 * bhuey waits for the Sprint update Jul 11 03:53:19 join #webos Jul 11 03:53:22 bah Jul 11 03:53:26 whoops :/ Jul 11 03:54:21 tman_mac: ok, well I'll try and review this and post something on the wiki. hopefully it'll help other people get started too. Jul 11 03:54:31 hborders: sounds great :) Jul 11 03:55:27 tman_mac: one last thing: is agar's use of opengl transparent to me as the consumer? If I write a custom control, does it have to use opengl commands to render or something? Or does agar have its own drawing apis that it translates into sdl or opengl? Jul 11 03:56:07 hborders: the latter. it's a pretty well-done library Jul 11 03:56:54 tman_mac: sah-weet. I'm surprised that palm doesn't just include agar with their pdk stuff Jul 11 03:57:26 tman_mac: really, I'm surprised palm doesn't just buy the widk and repackage it as pdk 2.0 Jul 11 03:57:39 lol Jul 11 03:57:48 widk is great Jul 11 03:58:01 yeah, I'm blown away by all the libs it has Jul 11 03:58:25 and I'm glad I was able to get one put in Jul 11 03:58:54 yup Jul 11 03:59:24 the whole collaborative OSS model really works for this situation :) Jul 11 04:01:20 hey everyone ... don't mean to interupt the conversation, but have an update to save/restore...anyone working on Preware? want to push an update Jul 11 04:03:10 guess that give me my answer! :-) Jul 11 04:03:58 heheh.. I'm not :p Jul 11 04:05:31 rod,you there? Jul 11 04:08:56 well,so it's in the logs, I was not able to replicate the NAN error,tried to reinstall the warthog-69 with gov5.6, all good this time, put F105,no problems at all either Jul 11 04:09:08 isolated issue I guess Jul 11 04:09:52 didn't we already know the "nan error" was because of jhos code in that one version? Jul 11 04:13:16 hey now Jul 11 04:13:27 it was 4am :p Jul 11 04:13:43 and I fixed it the next day Jul 11 04:13:49 pffft :) Jul 11 04:13:56 people went batshit crazy about it though Jul 11 04:13:59 for some reason Jul 11 04:14:04 I know Jul 11 04:14:04 they came outta the woodworks to point it out Jul 11 04:14:05 haha Jul 11 04:14:07 lol Jul 11 04:14:20 guys.. we are soon to have arch support in WIDK :) Jul 11 04:14:35 my answer is the same as bender's Jul 11 04:15:11 what, you're gonna bend it? Jul 11 04:15:14 arch as in the scm arch or multiplatform build? Jul 11 04:15:57 jhowirc: multiplatform. armv6 is working for me now. with some work on extracting the rootfs for i686, it could be added easily Jul 11 04:16:08 ah Jul 11 04:16:58 I think I have unwittingly set up most of widk for myself while trying to get kernels built Jul 11 04:17:09 bad news: when I merge to master, you all have to do a 'make clobber-armv7' :> Jul 11 04:17:27 unacceptable Jul 11 04:17:42 make make clobber invoke that Jul 11 04:17:46 :) Jul 11 04:17:56 well you can certainly do that Jul 11 04:18:07 but then you'd have to rebuild the toolchain, which isn't necessary Jul 11 04:18:40 there is already too much rebuilding so I agree Jul 11 04:18:51 my mac is about to take off.. make -j5 and the fan is going nuts :p Jul 11 04:19:17 yeah, all the svn packages have to rebuild every time for some reason Jul 11 04:19:34 I can't wait for 1.4.5 to get here so that we don't have to build kernels 4 times just to get 1 release out Jul 11 04:20:12 jhowirc: 4 times? Jul 11 04:21:10 yeah. 1.4.1 1.4.2 1.4.3 1.4.5 Jul 11 04:21:20 oh right Jul 11 04:21:21 Tman_mac: which mac? Jul 11 04:21:42 jacques: my macbook pro i5 Jul 11 04:21:43 man my sprint pre keyboard is creaking like and old man Jul 11 04:21:55 tman I'm jealous Jul 11 04:22:13 Tman_mac: ah cool, I'm considering getting one Jul 11 04:22:29 s/and/an/ Jul 11 04:22:35 yeah I love it.. the timing was right to get one so I did :) Jul 11 04:23:09 Tman_mac: are you running linux on it? Jul 11 04:23:53 jacques: in Parallels. linux is great but I <3 OSX Jul 11 04:24:33 ok I think that's how rwhitby runs it too. I would want to dual-boot Jul 11 04:25:13 whatever works for you. in fact with Parallels you can run a boot camp partition, so you could have the option to boot into it or virtualize it Jul 11 04:25:32 that's pretty nice Jul 11 04:25:36 I would use that Jul 11 04:25:37 indeed Jul 11 04:26:47 who's familiar with the workings of the WIDK build system? particuarly when it comes to these .depends.inc files.. Jul 11 04:27:13 I can see where they're generated, but not where they actually get included Jul 11 04:28:11 .depends.inc sounds like a new feature to me Jul 11 04:28:31 tman_mac: more sdl/agar questions: what is the advantage of SDL_PollEvent over SDL_WaitEvent? If there is an event in the queue, won't SDL_WaitEvent just return? Jul 11 04:28:35 (I haven't messed with WIDK in too long) Jul 11 04:30:31 hborders: yes it will. it depends on if you want for the event loop to wait for an event or not. mine doesn't; if there are no events it calls an update() function because it needs to be constantly updated and not wait for an event to update Jul 11 04:31:23 soo.. basically the event loop consists of handling any events that may or may not be waiting (inner loop) and calling update, without waiting (unless mPaused is true) Jul 11 04:31:31 yeah Jul 11 04:31:45 but that totally depends on the application Jul 11 04:31:54 so you might have your game logic running on the ui loop and thus you want it to keep spinning regardless Jul 11 04:32:08 is there a way to trigger sdl events from other threads? Jul 11 04:32:25 so I could wake up the event loop if I wanted to ask it to paint? Jul 11 04:32:56 possibly.. look in the SDL wiki under events Jul 11 04:34:13 in my case, I have a processing thread going that increments a number every time it updates.. the update code then keeps the number and compares it on the next call to determine if it needs to update Jul 11 04:35:01 maybe not the best way.. and a different approach might be better for other applications.. idk Jul 11 04:36:47 Tman_mac: what kind of app are you building? Jul 11 04:37:06 tman_mac: thanks. I'll make sure to put both into the example and explain it. Jul 11 04:38:03 jhowirc: real-time analyzer, analyzes sound data.. the process thread is what analyzes, and it increments the counter every time it is done with another period of sound data Jul 11 04:38:19 nice. Jul 11 04:38:20 and the result is then displayed Jul 11 04:38:29 it is quite intricate :) Jul 11 04:38:49 especially with the bajillion libraries it makes a use of.. Jul 11 05:08:02 Hi all... apparently I have 1.4.5 waiting. Anything special I need to do before upgrading? I'm running unixpsycho's 800 mhz kernel... Jul 11 05:53:57 moo Jul 11 05:54:05 sweet. bip works Jul 11 06:34:54 Does anyone know if Delta Dagger includes compcache stuff? Or only screenstate from Starfighter? Jul 11 06:35:05 Unix's web page doesn't help me Jul 11 06:42:22 Delta Dagger? These kernel names are starting to sound suspiciously like Debian/Ubuntu releases.. Jul 11 06:43:06 lol Jul 11 06:43:15 the names for the adult version of ubuntu are great Jul 11 06:47:22 So noone knows? :P Jul 11 08:12:16 hi Jul 11 08:14:38 is it possible to format the usb flash? Jul 11 08:15:13 it seems that my pre has a damanged filesystem Jul 11 08:15:52 hucksy: have you tried webos doctor? Jul 11 08:16:18 Or even a simple file system scan in windows Jul 11 08:17:35 nope, webos doctor dont touch the filesystem on the usb flash. I dont have windows but when its possible, i can make a mkfs on linux Jul 11 08:17:35 mhm Jul 11 08:52:26 lingfish: make sure your kernel comes from a preware feed Jul 11 08:54:50 hucksy: filesystem is checked on each boot Jul 11 08:55:13 oh ok Jul 11 08:55:13 if unrecoverable errors, the startup scripts wipe the partition Jul 11 08:56:35 or you can use meta-doctor to memboot and work on the unmounted partitions Jul 11 08:57:10 thanks! Jul 11 09:43:36 rwhitby: still there? Jul 11 09:53:55 hape: yes Jul 11 09:58:12 I send you a mail about the speechd question Jul 11 09:58:45 I whanted to ask you for your timing ideas to send navit to the autobuilder Jul 11 09:59:05 That I'm prepared to create a new thread in Precentral and watch there for questions Jul 11 09:59:35 hape: here is my proposal Jul 11 09:59:53 we create a git.../libraries/espeak.git Jul 11 10:00:19 in there, you store the source code, and the generated binaries, much like in (for instance) games/sdldoom.git Jul 11 10:00:37 we expect that the source code can be built with a Makefile in there that uses the WIDK setup Jul 11 10:01:03 hape: then build.git/nonworking/speechd/Makefile downloads it from there Jul 11 10:02:58 hape: is speechd going to have a mojo front end ? Jul 11 10:03:10 no speechd is just a service Jul 11 10:03:34 it could have a simple mojo front-end? type in words and hit a "speak" button? Jul 11 10:03:37 noradtux has the idea to add a webos service to the currnt socked based service Jul 11 10:03:59 Then it could be used by any WebOS appliaction so speak text Jul 11 10:04:01 yes, we would like to add a C service aspect to it Jul 11 10:04:55 at the moment not as a nojo app can not call it. we could only make it the way we did with espeak. call a script that speaks some predifined text Jul 11 10:06:45 We spoke about that and decidet not to do it. As it is only a service and would waist space on a allready filled up langer pages Jul 11 10:06:54 ok, but we should plan for a mojo service and application in the future Jul 11 10:07:44 mojo service is on noradtux plan Jul 11 10:08:07 as it is allready a running packground service demon. Jul 11 10:08:17 hape: espeak has a command line front-end, able to speak works from stdin, right? Jul 11 10:08:43 right. But we do not use espeak direktly right now Jul 11 10:09:01 what is the structure right now? Jul 11 10:09:17 there is a command line interface for speechd Jul 11 10:09:41 where is speechd from? Jul 11 10:09:42 speechd is the speed dispatcher. You send the text to him as fast as you what Jul 11 10:10:02 he the calles espeak as a lib to speak the text one after each other Jul 11 10:10:47 would we ever want one and not the other installed in the future ? Jul 11 10:11:16 hape: I have time now for the next couple of hours to work on this. do you? Jul 11 10:11:37 yes Jul 11 10:12:02 ok, what packages in WIDK are used? Jul 11 10:12:15 and do they work on the 'arch' branch? Jul 11 10:13:07 noradtux crated a preivate feed with your scripts Jul 11 10:13:13 http://tnx.homeip.net/~norad/myfeed/ Jul 11 10:13:35 So this are the pages as they needed in filnal Jul 11 10:13:36 now, a question about locale Jul 11 10:13:58 at the moment the makefile uses the loacle from the compiling host Jul 11 10:14:18 why does the symlink need to go through /media/internal? why not just symlink straight from /usr/lib/locale to the app dir in /media/cryptofs/apps/usr/palm/applications/org.webosinternals.locale/... ? Jul 11 10:14:31 yes Jul 11 10:15:02 The makefiles in the build git contains the needed pre/post scripts Jul 11 10:15:12 why does the symlink need to go through /media/internal? why not just symlink straight from /usr/lib/locale to the app dir in /media/cryptofs/apps/usr/palm/applications/org.webosinternals.locale/... ? Jul 11 10:16:27 The needed packages from the crosscompile git are apps/navit apps/speechd and common/locale Jul 11 10:16:56 The espeak package is part of the speechd package. The seperate espeak package is not used anymore Jul 11 10:17:37 what if someone else wants to use the espeak package without speechd? why not keep them as one package depending on another? Jul 11 10:18:10 rwhitby: the jail only allows access to the one appliaction folder in cryptos. I we link it to the locale folder in cryptfs then it is not seen in the jail Jul 11 10:18:28 to use the espeak package direktly is not supported Jul 11 10:18:55 speechd is the dispatcher frontend to get all requests Jul 11 10:19:45 if you call espeak faster as it can speak you get into trouble. speechd is taking care of this and more Jul 11 10:19:53 ok Jul 11 10:20:25 I see no reason to call espeak instead of speechd Jul 11 10:20:43 ok Jul 11 10:20:54 from the commandline you can call speexhd with the spd-say command in the seechd package fodler Jul 11 10:21:11 ok, that will work for the service then Jul 11 10:21:25 I can quickly make a mojo service and front-end using that Jul 11 10:21:47 and that package will be useful independent of navit, so it should have a good name too. Jul 11 10:23:00 H�. The plan is to extend the exisitin speechd demon to contain the service. Not another running deamon for the service Jul 11 10:23:02 is that package compatible with the app catalog content rules? Jul 11 10:23:33 not shure, never read the app catalog content rules Jul 11 10:23:50 I agree it should be one executable speechd and service. Jul 11 10:24:14 so we need a good name for it, cause "speechd' is not an end-user friendly name (no end user knows what a daemon is) Jul 11 10:24:44 the name will cover the daemon/service, and the mojo app front-end, all in one package. Jul 11 10:24:54 At the moment speechd can be called by his C API or wia a socked. Jul 11 10:25:12 The second way is how Navit calls speechd Jul 11 10:25:14 (just like Govnah is the name of the service and front-end, and same for the Preware package now) Jul 11 10:26:25 That is the home of speechd http://www.freebsoft.org/speechd Jul 11 10:28:05 should the espeak library be packaged separately, so that someone else can replace it with a festival back-end later if they wish? Jul 11 10:28:46 we not througt about that. But it could be done Jul 11 10:30:06 hape: is the locale library complete? does it have everything it will ever need in the future, or do other applications in the future write into the same directory? Jul 11 10:30:31 Not sure if we should create on more package. I think it would be better to replace the the speechd packge with a other one that containf the festival back-end Jul 11 10:31:04 and will the locale package be used by other applications other than navit? Jul 11 10:32:32 The locale can be used by other C applaictions that uses the libc for translations Jul 11 10:33:33 is there a way to do it without needing /usr/lib/locale? cause that will never be accepted into app catalog Jul 11 10:34:13 The only need to do export LD_LIBRARY_PATH=$APP_DIR/lib:/media/internal/.local/lib before the app ist started Jul 11 10:35:04 I expect yes, not sure why noradtux sill add this link Jul 11 10:35:31 the problem is that the path is hardcoded in the libc from plam. But the files are missing Jul 11 10:39:15 I think we need a more generic solution for the /media/internal/.local directory. something that can be used by many different applications for sharing information between jails Jul 11 10:39:31 the current locale packed has all we need i think. But we could send out another on later if a language is missing Jul 11 10:40:05 ok, I understand the locale stuff now, thanks. Jul 11 10:40:37 It costed noratux some time to find this all out. Jul 11 10:43:13 ok, we now have applications/navit.git services/speechd.git and libraries/locale.git Jul 11 10:43:37 for navit, please structure the repo like the applications/preware.git repo Jul 11 10:44:03 So the sepearte gits works like the seperate packge folders in the crosscompile git? Jul 11 10:44:57 no, they work as the source and binaries git for the autobuilder, with the Makefile in preware/build.git doing the packaging Jul 11 10:50:43 so we store the pregenerated binaries there? Jul 11 10:51:45 yes Jul 11 10:52:14 if they are built precisely from the source and makefile in WIDK, then we just need a README and LICENSE to point to the source in WIDK Jul 11 10:52:45 in the long term future, when cross-compile.git and build.git are somehow merged, this will not be necessary. Jul 11 10:52:56 ok, I expected the autobuilder would do the compiling to. So I was worng Jul 11 10:53:20 for simple things like preware and govnah and saverestore and wIRC, it does. Jul 11 10:53:30 but the autobuilder does not autobuilder WIDK yet Jul 11 10:53:49 but the plan is that it will. so these repos are a temporary measure until that happens. Jul 11 10:54:08 ok, got it Jul 11 10:56:05 so we need to compile the packages then and put the resuts in the binary git repositories Jul 11 10:56:44 so the locales tar.gz is 1.7MB? Jul 11 10:57:31 ah, LC_COLLATE and LC_CTYPE are large Jul 11 11:01:06 yep they are Jul 11 11:01:44 hape: would there be an objection to renaming /media/internal/.local to another name under /media/internal ? Jul 11 11:01:54 that is why we try to include only the locale for the langs that are supportet by Palm Jul 11 11:02:06 yeah, it's a good set Jul 11 11:02:28 hape: cause we will want to use the same directory for the X server applications too Jul 11 11:02:35 Not problem from my side. We only should keep it hidden for the normal users. That the don't mess it up Jul 11 11:02:37 and the Qt ones too Jul 11 11:02:45 yes, it will be hidden Jul 11 11:02:59 bah Jul 11 11:03:26 any new developments going to hit any uberkernels soon ? Jul 11 11:03:36 nope Jul 11 11:04:20 rwhitby: question on kenal. The 1.4.5 uberkernal is still in testing or? Jul 11 11:04:36 hape: it will stay in testing until the 1.4.5 kernel source is released by palm Jul 11 11:04:50 since it is actually a 1.4.1 kernel, just packaged for installation on 1.4.5 Jul 11 11:05:44 hape: so this directory on /media/internal/. will be used by all widk applications that need to share data outside the jail Jul 11 11:06:19 so it needs to have some structure underneath it to avoid collisions and allow for easy installation and removal of items Jul 11 11:07:10 makes sense Jul 11 11:08:08 and we need to realise that it is not accessible (i.e. an access will cause an IO error) when the USB drive is active Jul 11 11:08:26 we could take it as the folder inside /. would be a root folder or? Jul 11 11:09:13 you are right. Never thourgt about that. Normaly I quit all apps before I start USB moder Jul 11 11:09:17 you are right. Never thourgt about that. Normaly I quit all apps before I start USB mode Jul 11 11:09:26 well, there are two ways to structure. by appid or by mirroring structure under / Jul 11 11:09:35 or by both (structure under appid) Jul 11 11:10:43 we also need to make sure that postinst scripts can handle being run twice, and also being run after a partial erase that removed appdir but did not remove /media/internal/.name Jul 11 11:11:35 and it should be a name that will not conflict with a name that Palm might want to use in the future Jul 11 11:11:38 is 1.4.5 out yet ? are you just prepping packages for that release ? Jul 11 11:11:51 bhuey: 1.4.5 is out in all major countries Jul 11 11:11:59 but not in the USA yet ;) Jul 11 11:12:00 :/ Jul 11 11:12:24 rwhitby: 720mhz and 24M compcache are a significant improvement for me minus the lack of storage on my original Pre Jul 11 11:12:38 thanks to you and the crew for this improvement Jul 11 11:12:46 rwhitby: yeah, it's a pain Jul 11 11:12:53 bhuey: we simply build on the kernel developers before us Jul 11 11:13:08 I wish it was released yet. Has the elimination of the TMC warning been confirmed ? Jul 11 11:13:19 lol Jul 11 11:13:36 or was it somehow delaying the inevitable and the bug is just masked Jul 11 11:13:37 ? Jul 11 11:14:02 rwhitby: still, that's a great service because me and others don't have the time to do that kind of work Jul 11 11:14:12 bhuey: I think there was probably just a memory leak somewhere in LunaSysMgr or LunaBrowserService and they've probably just fixed it Jul 11 11:14:58 I just turned the uberkernel "on" to some random guy that we were hanging out with and it impressed him Jul 11 11:15:03 rwhitby: yeah, that's a major bug Jul 11 11:15:21 i'll know for sure by monitoring govnah and the amount of swap it uses Jul 11 11:15:29 hape: so I'm thinking /media/internal/.widk/ for the name Jul 11 11:15:30 if the bugs is fixed Jul 11 11:15:41 rwhitby: fine with me Jul 11 11:15:54 now for the structure under it Jul 11 11:16:04 dtzWill: ping Jul 11 11:19:18 rwhitby: so it's confirmed that the leak is fixed ? Jul 11 11:19:26 bhuey: I don't know Jul 11 11:19:36 kind of odd that they didn't make a mention about the announcement leaks Jul 11 11:19:42 I don't even know if there was a leak or not Jul 11 11:19:44 yeah, so it's still not been confirmed Jul 11 11:20:04 defintely a leak, didn't try to track it down though Jul 11 11:20:45 I've restarted both luna and java, the leak persists Jul 11 11:20:59 I know that the TMC is based on (used ram plus used swap) / (total ram) and the omission of total swap from the denominator was intentional by Palm to keep the phone out of swap. Jul 11 11:21:06 hi Jul 11 11:21:10 noradtux: hi Jul 11 11:21:24 just reading the chat-log Jul 11 11:21:34 rwhitby: it eats swap over time Jul 11 11:21:38 noradtux: I have time tonight to get Navit sorted out into Preware Jul 11 11:21:55 noradtux: any comments or objections on the discussion so far? Jul 11 11:21:59 I get sound stutter with this uberkernel but it's worth it Jul 11 11:22:33 I am at the part where you ask if it makes sence to keep espeka as a separate package Jul 11 11:23:13 noradtux: if needed, we can split speechd into two packages with a dependency in the future, and preware will handle that gracefully on the upgrade Jul 11 11:23:33 rwhitby: well, that is an option Jul 11 11:23:34 so there is no need to split it now just for a potential swappable back-end in the future Jul 11 11:24:11 and that confirms that the package be named by speechd rather than be named by espeak Jul 11 11:24:51 noradtux: is there any way to have locale files in /usr/local/lib/locale ? Jul 11 11:24:53 rwhitby: ok, so let's keep espeak and speechd separate Jul 11 11:25:34 hmm, they are about 10 megs, my pre has just 24MB free there, so I do not think that is an option Jul 11 11:25:38 noradtux: I was thinking to keep them in a single package until someone decides in the future that they want to swap out the backend Jul 11 11:25:55 noradtux: org.webosinternals.widk does a bind mount for /usr/local Jul 11 11:26:09 rwhitby: it does? Jul 11 11:26:29 where/how Jul 11 11:27:02 noradtux: yes, the idea is that you can compile apps and libraries using the WIDK to install into /usr/local/... and then it is bind mounted by an org.webosinternals.widk dependency bootstrap package Jul 11 11:27:56 however, /usr/local is not in the jail configuration, so that has killed that dead Jul 11 11:28:49 noradtux: another idea I have had (and have implemented in meta-doctor) is to add a new certification authority to the signature verification chain, and then use the per-app jail configuration support Jul 11 11:30:26 then an app could specify in it's jail_app.conf file that it needs to access other app dirs Jul 11 11:31:38 noradtux: at the moment, the only package that uses the widk /usr/local mount is the gsmwav package Jul 11 11:32:04 Isen't there a risk in chnaing the way that the jail is working if Palm makes chnages to this later too? Jul 11 11:33:58 rwhitby: oh, so there _is_ the possibility to modify the jail on a per-package level? Jul 11 11:34:33 noradtux: I have not tested it yet, but in /usr/bin/jailer I see the signs of signature verification and jail_app.conf Jul 11 11:34:39 (btw. I've read the log now :) Jul 11 11:34:56 so I am assuming such a possibility at the moment Jul 11 11:35:11 ah, I see Jul 11 11:35:13 "Switching to validated app conf file %s" Jul 11 11:35:24 "App conf file %s signature does not validate." Jul 11 11:36:00 my guess is that jail_app.conf and jail_app.conf.sig work in the same way as normal Palm package signatures, using /etc/ssl/certs/appsigning-bundle.crt Jul 11 11:36:42 and I already have that working in meta-doctor where it makes Palm install a signed Preware package on first boot over the network Jul 11 11:36:56 (the signature part, not the jail part) Jul 11 11:37:57 do you modify that mundle? Jul 11 11:38:00 bundkle Jul 11 11:38:03 bundle Jul 11 11:38:08 yes, I add the Preware CA to it Jul 11 11:38:36 meta-doctor does that at the moment, but Preware install script could easily do it too Jul 11 11:39:28 so then anything signed by the preware.org autobuilder could have an application-specific jail configuration file Jul 11 11:39:58 can't we just append the certificats we need to that file? Jul 11 11:40:13 it needs to be a CA, not just a cert. Jul 11 11:40:48 wait, I misunderstood that statement about preware could be alble to do it too. Jul 11 11:41:02 the single preware postinst script could do it Jul 11 11:41:20 so that any packages that go through the autobuilder are signed with a cert that is then valid Jul 11 11:41:28 yes, that's what I thought Jul 11 11:42:55 so the question is whether we want to just do a /media/internal/.widk/usr/lib/locale solution, or whether we want to try for Preware-signed custom jails. Jul 11 11:43:36 well, the hidden directory would be easier Jul 11 11:44:03 and waht is the saver verion if we think about later changs of WebOS by Palm Jul 11 11:44:13 yes, and we can always seamlessly transition to the latter in the future if desired Jul 11 11:44:36 hape: well, Palm could always change anything Jul 11 11:45:25 but in the past our criteria has been to try and use the things that Palm provides in the way that they are meant to be used, without breaking any security while doing so. Jul 11 11:45:33 yes, basically you are screwed as soon as you leave the path (APIs) that palm provides you Jul 11 11:46:58 and since it seems that signed app-specific jail configurations are part of /usr/bin/jailer (a proprietary Palm creation) then one must assume they are going to use that to allow them to sign app configurations on a case-by-case basis for authors (e.g. if you submit two apps that need to work together) Jul 11 11:47:25 cause Palm can just give the jail_app.conf and jail_app.conf.sig files to the developer to be packaged up as part of the ipk Jul 11 11:47:36 yes Jul 11 11:48:01 got it Jul 11 11:48:05 but I agree that /media/internal/.widk/usr/lib/locale is the simpler solution for the short term Jul 11 11:48:22 I'd like to see jail_app.conf used somewhere before depending on it. Jul 11 11:48:47 noradtux: so, do you want to commit the current binaries to the repos I have created, and I will set up the autobuilder to package from there? Jul 11 11:48:50 agreed. btw another candidate to put into /media/internal/.widk/usr/lib/ would be libspeechd.so.2 Jul 11 11:49:08 yes, after lunch ;) Jul 11 11:49:33 or do you want me to put them in there in the structure I would prefer? Jul 11 11:49:37 I can confirm that the binaries work as they are now Jul 11 11:50:01 from the packages on your temporary feed Jul 11 11:50:06 hmm, yes, I think that would be best. I don't exactly know how you would like it Jul 11 11:50:12 yes Jul 11 11:50:16 those packages work Jul 11 11:50:26 ok, I'll do that, and change the .local path at the same time Jul 11 11:50:31 ok, i'm off, eating :) Jul 11 11:50:41 noradtux: Guten Appetit Jul 11 11:52:00 noradtux|away: what is pdl-helper? Jul 11 11:53:11 is reads the current language setting with a PDL api call and retus it as variable Jul 11 11:53:33 so we can set the needed exports for speech lang and Navit lang Jul 11 11:54:30 http://git.webos-internals.org/?p=preware/cross-compile.git;a=blob;f=packages/apps/navit/files/pdl-helper.c Jul 11 11:54:33 is the /usr/lib/locale symlink actually needed? Jul 11 11:54:51 or is the LD_LIBRARY_PATH to /media/internal/.widk/usr/lib enough? Jul 11 11:54:56 Palm defined a new data path for PDL apps Jul 11 11:55:16 oh, do we know where that ends up Jul 11 11:55:18 ? Jul 11 11:55:40 I think the link to /usr/lib/locale is not needed, but i diden't tested it Jul 11 11:55:52 ok, I will package without it to start Jul 11 11:56:01 the less changes to Palm directories, the better. Jul 11 11:56:16 in /media/internal/appdata/org.webosinternals.navit Jul 11 11:56:34 really? /appdata, not /.appdata? Jul 11 11:56:46 that is what we get from PDL_GetDataFilePath("", buf, BUFSIZE); Jul 11 11:57:14 and it definitely does not have a dot ? Jul 11 11:57:35 yes Jul 11 11:57:48