**** BEGIN LOGGING AT Sat Apr 15 09:59:57 2006 Apr 15 10:17:27 damn, now I need zecke's help again Apr 15 10:19:52 mithro: Anything I can help with? Apr 15 10:20:03 * RP is hacking on the secondary cache... Apr 15 10:21:39 i need to know initalise the C parser Apr 15 10:22:11 You need zecke :-(. I never did manage that myself :-/ Apr 15 10:22:29 whats the secondary cache do? Apr 15 10:23:19 What I mentioned the other day - create a small subset of cache data instead of caching the whole parsed data Apr 15 10:23:43 So it will only speed up reparsing where nothing has changed but that's a good improvement to have Apr 15 10:25:22 RP: have you tried running the code through the python profiler? Apr 15 10:25:41 mithro: zecke did and was working from that Apr 15 10:26:11 mithro: I know currently bitbake spends 33% of its time doing disk IO writing out the cache on my machine Apr 15 10:26:45 why is the cache so big btw? Apr 15 10:27:20 mithro: When all the classes combine, we have a large amount of variables Apr 15 10:29:45 still seems a bit excessive Apr 15 10:30:17 If you can see any improvements, we'd take them... Apr 15 10:30:24 400MB of cached metadata to fill 8MB of flash Apr 15 10:30:47 mithro: Each cache file is ~100kb which isn't so bad... Apr 15 10:31:17 bitbake code makes my head hurt :P Apr 15 10:32:28 I didn't know where to start yesterday. I'm slowly getting somewhere now though. I can generate the list of variables I need to use a secondary cache for Apr 15 10:37:40 dunno how anyone wrote python code like this :P Apr 15 10:38:30 mithro: Its a nightmare although my python is limited to start with :-/ Apr 15 10:48:12 morning Apr 15 10:52:17 hi CoreDump|home Apr 15 10:52:41 CoreDump|home: I found that dd bug btw - there's a fix on LKML Apr 15 10:52:48 nice! Apr 15 10:53:02 Well, several fixes - I just need to confirm it works Apr 15 10:53:32 if I can be of help let me know Apr 15 10:54:14 CoreDump|home: Will do. Hopefully I'll push a patch into OE soon. I want to get this bitbake work done first :) Apr 15 10:54:30 no hurry =) Apr 15 10:58:49 RP: c7x0 was moved to kernel 2.6 right? Apr 15 10:59:21 CoreDump|home: Correct - it was the first Apr 15 10:59:40 excellent =) Apr 15 10:59:54 * CoreDump|home drops kernel 2.4 xmodmap for c7x0 Apr 15 11:00:47 ~sheperd Apr 15 11:01:06 ~shepherd Apr 15 11:01:07 [shepherd] a Sharp Zaurus SL-C750, or a dog. Apr 15 11:01:21 ~husky Apr 15 11:01:23 somebody said husky was Sharp Zaurus SL-C760 or a dog Apr 15 11:01:24 ~corgi Apr 15 11:01:26 corgi is, like, sharp sl-c700, or a dog Apr 15 11:03:46 It appears I just regressed bitbake's memory usage to the bad old days :-/ Apr 15 11:04:19 RP: If you made it as fast as it was back then I could live with id =D Apr 15 11:05:31 CoreDump|home: We had 10% of the .bb files we have now :-/ Apr 15 11:06:27 we had like 1700 Apr 15 11:06:51 http://hentges.net/tmp/screenshots/mhcln03/oe.jpg Apr 15 11:06:56 2137 =) Apr 15 11:06:59 CoreDump|home: And we're slower than we were then? Apr 15 11:07:29 Yep, at least on my box the initial parsing takes way longer Apr 15 11:07:34 (now) Apr 15 11:08:00 It'd be nice to know why as that's worrying Apr 15 11:08:01 hi zecke Apr 15 11:09:00 moin Apr 15 11:09:09 hey zecke Apr 15 11:10:59 CoreDump|home: it can't take longer, use the kernel you used before Apr 15 11:11:23 CoreDump|home: have you had some special attributes in your fstab? (no atime or such?) Apr 15 11:11:37 zecke nope Apr 15 11:11:56 I have sinced moved to a XFS RAID with noatime and friends Apr 15 11:12:21 and I really do not want to use kernel 2.4.20-bk36 anymore ;) Apr 15 11:13:10 CoreDump|home: I refuse to consider that bitbake got slower ;) Apr 15 11:13:21 CoreDump|home: "Was nicht sein darf, kann nicht sein" Apr 15 11:13:22 zecke: I know =) Apr 15 11:13:33 you told me that shortly after the change heh Apr 15 11:14:01 CoreDump|home: turn off the cache please (CACHE="") Apr 15 11:14:12 CoreDump|home: we will see quite some memory consumption then but... Apr 15 11:14:25 I got 2G in that box Apr 15 11:14:35 in local.conf or shell env? Apr 15 11:15:23 local.conf Apr 15 11:15:27 time bitbake foo Apr 15 11:15:27 NOTE: Using cache in '/usr/oe/build2/tmp/cache' Apr 15 11:15:28 NOTE: Parsing finished. 3206 cached, 0 parsed, 66 skipped, 0 masked. Apr 15 11:15:28 ERROR: Nothing provides dependency foo Apr 15 11:15:29 real 0m6.599s Apr 15 11:15:29 user 0m5.692s Apr 15 11:15:30 sys 0m0.196s Apr 15 11:15:46 RP: ? Apr 15 11:15:57 zecke: secondary caching :) Apr 15 11:16:16 CoreDump|home: please run time bitbake -p opie-image with the fast and slower version Apr 15 11:16:28 CoreDump|home: I would like to know where the difference is spent Apr 15 11:16:36 will do Apr 15 11:16:41 RP: hehe, did you push that change yet? Apr 15 11:16:49 * zecke has big trouble with svk ATM Apr 15 11:16:53 zecke: No, it only just started to work :) Apr 15 11:17:03 zecke: At the moment it will crash if it builds anything as well ;-) Apr 15 11:17:07 It forgot from which depot it copied bitbake-tests Apr 15 11:17:12 zecke: But this is a minor issue ;-) Apr 15 11:17:24 RP: oh well, you should have shipped it more early Apr 15 11:17:31 anyone here who can send me a modified hciattach binary? Apr 15 11:17:47 modified? Apr 15 11:17:48 RP: This diverts from the breakage I'm creating Apr 15 11:18:11 goxboxlive: e.g. replacing the ELF header with MCOF? Sadly not Apr 15 11:18:13 zecke: I know and this worries me :-/ Apr 15 11:18:22 ah Apr 15 11:18:47 zecke: I had to back out the changes between r419 and r423 to get something I could work on as I can't get r423 to work properly Apr 15 11:19:04 RP: hmm, that is weird Apr 15 11:19:37 zecke: It doesn't like the DEPENDS_prepend in package_ipk Apr 15 11:20:46 I was seeing -native packages depending on ipkg despite the fact native packages have PACKAGES="" and therefore don't pull in ipkg Apr 15 11:20:47 I want the one who is able to use caption s '-S' The one in the familiar-0.8.4.rc1 doesnt support caption s. Apr 15 11:21:05 RP: okay, that is a hint ;) Apr 15 11:21:16 RP: regardles I'm passing all the tests :} Apr 15 11:21:32 zecke: :-/ Apr 15 11:21:44 That shows off the quality of them... Apr 15 11:22:03 zecke: It worked fine for a while, then broke. Nothing I could do (wiping tmp/cache etc.) would unbreak it... Apr 15 11:23:34 real 5m27.019s Apr 15 11:23:34 user 5m1.963s Apr 15 11:23:35 sys 0m7.164s Apr 15 11:23:43 that is with # CACHE Apr 15 11:24:41 hey zecke! Apr 15 11:24:50 how did you sneek in? Apr 15 11:25:15 mithro: from this small terminal over here Apr 15 11:25:24 i need your help again Apr 15 11:25:39 i can't figureout how to setup and call the C side of the parser Apr 15 11:29:36 RP: At least your statement correlates to the revisions you mentioned :) Apr 15 11:30:01 zecke: :) Apr 15 11:30:09 mithro: see bitbakescanner.l Apr 15 11:30:14 where? Apr 15 11:30:16 mithro: the 'parse' method Apr 15 11:30:25 mithro: lib/bb/parse/parse_c/bitbakescanner.l Apr 15 11:30:28 real 6m16.146s Apr 15 11:30:28 user 5m41.629s Apr 15 11:30:28 sys 0m10.679s Apr 15 11:30:31 this is with cache Apr 15 11:30:45 ahh I see it Apr 15 11:31:06 mithro: there are some bits missing though Apr 15 11:31:29 mithro: e.g. YY_INPUT should not be defined, instead the 'FILE*' should be set as input Apr 15 11:31:40 hrm? Apr 15 11:32:03 hey zecke Apr 15 11:32:08 mithro: this scanner used to operate on 'mmaped' files, and needed a custom read buffer method Apr 15 11:33:16 so um.... Apr 15 11:33:26 i thought the C side worked? Apr 15 11:33:33 mithro: sort of ;) Apr 15 11:33:42 mithro: the grammar and scanner is correct Apr 15 11:33:49 so what will happen when I call the parse function? Apr 15 11:35:26 mithro: lex_t::input needs a implementation Apr 15 11:35:43 mithro: alternatively we can make the scanner work directly on our FILE* Apr 15 11:36:33 should I be implimenting "void parse (FILE* file, PyObject* data)" instead of you? Apr 15 11:37:11 mithro: no parse is already implemented Apr 15 11:37:24 mithro: it setups the scanner and grammar, and starts parsing Apr 15 11:37:34 mithro: which will call into the e_* methods and it is reentrant Apr 15 11:38:25 RP: could we look into the DEPENDS_prepend issue for a moment? Apr 15 11:39:18 okay then Apr 15 11:39:26 what does reentrant mean? Apr 15 11:40:17 zecke: I'll need 5 minutes to get a working bitbake Apr 15 11:49:44 zecke: no ICQ? Apr 15 12:01:18 zecke: ok, I have bitbake producing this error - what debug info can I provide? Apr 15 12:02:10 RP: local conf, bb files etc. for a later test case ;) Apr 15 12:02:23 RP: is it failing with bitbake -b as well? Apr 15 12:02:23 03coredump 07org.oe.oz354x * ra6d6b0a4... 10/packages/xserver-common/ (3 files in 2 dirs): xserver-common: Remove xmodmap for kernel 2.4 on c7x0 and cxxxx Apr 15 12:02:31 zecke: No, -b works Apr 15 12:02:50 RP: okay that is bad then ;) Apr 15 12:03:31 ~fishslap zecke Apr 15 12:03:32 * ibot slaps zecke up side the head with a wet fish. Apr 15 12:04:03 CoreDump|home: BTW: Use reiser for better bitbake performance not that crappy XFS ;) Apr 15 12:04:25 zecke: The error I see is: http://pastebin.com/661308 Apr 15 12:04:45 zecke: over my dead body Apr 15 12:05:14 CoreDump|home: then live with the bad performance ;) Apr 15 12:05:26 CoreDump|home: using a fs that can not be shrinked is... Apr 15 12:05:49 * CoreDump|home prefers bad performance over a POS data-killing FS any day ;) Apr 15 12:06:07 RP: do you have an idea what is going wrong? Apr 15 12:06:12 zecke: I do not need to shrink / rezise my raid heh Apr 15 12:06:19 CoreDump|home: you're forgetting 'changes on disk format more that hans reiser changes socks' Apr 15 12:06:50 koen: =) Apr 15 12:07:12 I take the slow ext3 over reiser any day hehe Apr 15 12:07:14 RP: wait a second please Apr 15 12:07:26 But I am pleased so far w/ XFS Apr 15 12:08:19 CoreDump|home: I'm using HFS+ (journaled) and UFS at the moment Apr 15 12:08:31 CoreDump|home: UFS is not journaled but I have background filesystem checks Apr 15 12:08:41 CoreDump|home: Until it east your files. But ext3 does that for me too... Apr 15 12:08:50 zecke: Can't comment on the fancy Mac stuff =) Apr 15 12:08:53 s/east/eats/ Apr 15 12:09:02 CoreDump|home: UFS, Unix File System ;) Apr 15 12:09:05 never had any problems w/ ext3 Apr 15 12:10:03 bleh Apr 15 12:10:06 RP: well. update_data is likely to be the problem Apr 15 12:10:11 one step forward and one step back Apr 15 12:10:17 RP: it used to fork the following way: Apr 15 12:10:33 RP: for each key -> for each override -> apply_override Apr 15 12:10:49 RP: now we do for each override -> find key -> apply_override Apr 15 12:10:59 CoreDump|home: I think I solved some crashes when using the clearlooks theme Apr 15 12:11:13 03koen 07org.oe.dev * re7b98b54... 10/packages/gtk-engines/gtk-engines_2.7.4.bb: Apr 15 12:11:13 gtk-engines: add 2.7.4 Apr 15 12:11:13 * this fixes a lot of crashes with the clearlooks theme Apr 15 12:11:13 * it now uses cairo do draw lots of stuff, so it might be slower as 2.6.x on FPU less systems Apr 15 12:11:20 yay! Apr 15 12:11:39 slower yes, but it is written in C so it is faster again :} Apr 15 12:11:41 CoreDump|home: but it problably is a bit slower due to cairo Apr 15 12:12:13 C the gurantee for small, maintainable and fast applications Apr 15 12:12:17 zecke: it uses git, so it can't be slow! Apr 15 12:12:23 could you push that to .oz354x as well please? Apr 15 12:12:46 CoreDump|home: not without updating to gtk 2.8 Apr 15 12:13:09 hmm ok, probably not a good idea =) Apr 15 12:13:51 zecke: http://pastebin.com/661321 Apr 15 12:14:01 RP: OT: I have found an issue with FOO_${DOO} yesterday... Apr 15 12:14:09 zecke: The problem is when bitbake sees DEPENDS, its not = "" Apr 15 12:15:11 zecke: I think the evaluation of DEPENDS=${@autotools_dep_prepend(d)}${@["ipkg-utils-native ", ""][(bb.data.getVar('PACKAGES', d, 1) == '')]}${PACKAGE_DEPENDS} ${@base_dep_prepend(d)} is failing somewhere along the line Apr 15 12:18:46 hmm Apr 15 12:19:10 zecke: Although it works with bitbake -b Apr 15 12:19:41 RP: it ignore DEPENDS though ;) Apr 15 12:20:56 zecke: It might ignore it but it does find the right value... Apr 15 12:21:25 unlike when bitbake it calculating the dependencies which it finds ipkg-utils-native in there... Apr 15 12:22:22 doh, why is "wlan0" back Apr 15 12:22:43 anyone know how to list symbols in a file? Apr 15 12:22:51 CoreDump|home: You probably changed from hostap to orinoco and back Apr 15 12:22:56 mithro: nm, objdump? Apr 15 12:23:00 mithro: objdump? Apr 15 12:23:02 just an ipkg upgrade Apr 15 12:23:23 CoreDump|home: Perhaps hostap wasn't installed or something? Apr 15 12:23:31 I'm going to have to break bbshell's peek and poke commands to get the speed increase... Apr 15 12:23:47 RP: I've seen reports that kernel-module-ieee80211 change eth0 to wlan0 Apr 15 12:24:07 no clue, worked fine with "eth0" when I flashed it. An upgrade brings back wlan0 which doesn't get setup Apr 15 12:24:11 koen: Ah, that would explain things... Apr 15 12:24:13 RP: hmm could you send me your local.conf please? Apr 15 12:26:05 mehh hostap depends on that module Apr 15 12:27:14 zecke: Did you get the PM? Apr 15 12:28:25 RP: is it failing without collections as well? Apr 15 12:30:44 zecke: checking Apr 15 12:34:36 zecke: Same problem Apr 15 12:34:54 RP: good Apr 15 12:35:12 The pdaxrom people (person) are (is) beginning to irritate me :-/ Apr 15 12:35:33 RP: what have they done now? Apr 15 12:35:48 2.6 sucks, it can't do X. Our users will hate it if it can't do Y Apr 15 12:35:56 Just generally annoying me... Apr 15 12:36:13 they still haven't figured out alsa? Apr 15 12:36:28 koen: They haven't got that far yet ;-) Apr 15 12:36:41 that ieee thingy breaks WEP Apr 15 12:36:50 it's funny how there superior buildsystem can't handle 2.6 Apr 15 12:36:57 CoreDump|home: you need some extra modules for that Apr 15 12:37:07 kernel-module-arc4-2.6? Apr 15 12:37:08 koen: The lack of cpufreq, QVGA, video overlay and probably other things I've forgotten are upsetting them Apr 15 12:37:25 RP: ah, you mean on cxxxx Apr 15 12:37:36 koen: yes Apr 15 12:37:51 RP: they don't have to use 2.6 Apr 15 12:37:59 koen: Apparently ewi is also "always down" Apr 15 12:38:06 heh Apr 15 12:38:17 koen@bitbake:~/OE/monotone/org.openembedded.dev/packages/xlibs$ uptime 14:49:51 up 87 days, 3:53, 11 users, load average: 1.15, 0.87, 1.38 Apr 15 12:38:33 not for the past 87 days Apr 15 12:38:50 RP: and if you use the shell parse and peek DEPENDS Apr 15 12:38:54 RP: what do you get? Apr 15 12:39:00 koen: Addmitedly it did have a blip in network access recently but I'd not call that "always down" Apr 15 12:39:18 zecke: I can't try that easily :-/ Apr 15 12:39:20 RP: ewi is serving feeds for the uni project as well Apr 15 12:39:32 RP: so a lot of people get upset when its down Apr 15 12:39:47 koen: I have no problem with it ;-) Apr 15 12:39:54 RP: but what do they need from ewi? Apr 15 12:39:58 hail zecke Apr 15 12:40:24 koen: I told them to grab a snapshot of OE to get the bits they want instead of pestering me Apr 15 12:40:59 zecke: Can you reproduce this problem I'm seeing? Apr 15 12:41:08 grrr Apr 15 12:41:18 RP: not yet :} Apr 15 12:42:40 RP: add them to your ignore list as move on Apr 15 12:42:50 koen: monotone.nslu2-linux.org is not too far behind ewi :-) 12:42:01 up 72 days, 13:35, 1 user, load average: 0.11, 0.21, 0.33 Apr 15 12:43:16 Has anybody ever considered flashing bootloader instead of kernel on C3x00? Apr 15 12:43:28 rwhitby: nice Apr 15 12:43:45 koen: (it's72 days since that box was installed at OSUOSL :-) Apr 15 12:46:14 koen: If I totally ignore them, OZ/OE will no doubt get slated for lack of cooperation Apr 15 12:46:53 RP: if it keeps you saner, so be it Apr 15 12:47:12 que? Apr 15 12:47:12 RP: I need to pull first... Apr 15 12:47:20 why does e0image wants to build gcc4? Apr 15 12:48:27 koen: I also have this totally mad idea they might generate some useful 2.6 patches. I think I'm probably dreaming though Apr 15 12:49:22 zecke: r423 also ouputs a number of parsing errors you should probably look into :-/ Apr 15 12:49:22 RP: I haven't seen any pdaX patch that wasn't horribly armv5te/zaurus specific Apr 15 12:49:53 koen: I can imagine :-/ Apr 15 12:49:53 the arc4 modules works, but it is not a dependency and is not automatically loaded Apr 15 12:51:08 zecke: I'm going to have to go back to r419 to do any further work on the cache I was working on... Apr 15 12:51:18 CoreDump|home: hrw had a solution for that Apr 15 12:51:24 RP: sure Apr 15 12:51:29 CoreDump|home: he proposed adding the depends by hand to kernel.bbclass Apr 15 12:51:57 sounds "interesting" ;) Apr 15 12:53:39 * RP -> food Apr 15 13:01:40 RP: you are using a pre alpha distribution ;) Apr 15 13:04:51 RP: which parse errors are you talking about? Apr 15 13:05:01 RP: the one I mailed yesterday? Apr 15 13:07:23 okay Apr 15 13:07:47 I'm studying now Apr 15 13:21:41 Hello guys Apr 15 13:44:47 any C guys about? Apr 15 13:45:41 mithro: I'd ask the question... Apr 15 13:46:19 I have a .o with a function and a .o which needs the function but they won't link together Apr 15 13:46:45 What does the linker say? Apr 15 13:47:11 nothing Apr 15 13:47:17 as I'm compiling to a shared library Apr 15 13:48:48 :-/. Does the function show up twice in the shared library? Apr 15 13:49:10 If so, its not linking properly or it thinks tkey don't match for some reason Apr 15 13:49:13 how do I test that? Apr 15 13:49:27 objdump or nm should show the symbols Apr 15 13:51:30 00005e89 t e_postcat Apr 15 13:51:30 U e_postcat Apr 15 13:52:33 What commandline are you using to link the files? Apr 15 13:52:52 g++ -shared -fPIC bitbakeparser.o bitbakec.o -o bitbakec.so Apr 15 13:53:28 Its not a C/C++ difference is it? Apr 15 13:53:42 could be Apr 15 13:53:54 They have different calling conventions Apr 15 13:54:17 You could try forcing the function to be a C function if its in C++ code Apr 15 13:54:21 i have done 'extern "C" {' Apr 15 13:54:41 around the apprioprate functions Apr 15 13:57:01 and the callers knows this function as 'c' as well? Apr 15 13:57:09 (look so though) Apr 15 13:57:13 yes Apr 15 13:57:31 mithro: what happens if you try to link bitbakec.o first? Apr 15 13:57:36 same Apr 15 13:58:00 * mithro ponders ordering a pizza Apr 15 14:02:02 is there way to check the calling thing? Apr 15 14:02:47 | Apr 15 14:04:20 do these differ? Apr 15 14:04:29 static void e_postcat(lex_t (*__pyx_v_c),char (*__pyx_v_key),char (*__pyx_v_what)) { Apr 15 14:04:40 extern void e_postcat(lex_t*, const char*, const char*); Apr 15 14:04:45 is it the const? Apr 15 14:05:13 mithro: static? Apr 15 14:05:30 mithro: If its declared static, nothing will see it Apr 15 14:05:39 o... Apr 15 14:05:44 why? Apr 15 14:05:56 static means its not referred to outside that file Apr 15 14:08:56 oo... thats almost fixed it Apr 15 14:11:51 okay, why don't these match then :P Apr 15 14:11:53 void parse (FILE* file, PyObject* data) Apr 15 14:11:53 void (parse(FILE (*),PyObject *)); /*proto*/ Apr 15 14:13:43 I'm not sure about the void() - what happens if you make them match a little more closely? Apr 15 14:14:00 ie void parse(FILE (*),PyObject *); Apr 15 14:15:43 ahh thats fixed now Apr 15 14:18:17 http://www.mi.fu-berlin.de/kvv/?veranstaltung=1133 <- I feel wasted at Uni... Apr 15 14:21:54 lex_t::input(char*, int*, int) Apr 15 14:22:03 was that what needed to be implimented? Apr 15 14:23:45 mithro: it could be completely dropped if we make the scanner/lexer use our FILE* as YY_INPUT Apr 15 14:24:12 mmap is most probably better? Apr 15 14:24:30 mithro: my idea was to open the file in python Apr 15 14:24:42 why? Apr 15 14:25:17 mithro: I wanted to avoid string conversion issues, raising exceptions from within c... Apr 15 14:25:26 string conversion issues? Apr 15 14:25:54 mithro: getting the char* from the Py_Object* ;) Apr 15 14:26:04 zecke: no problem there :P Apr 15 14:26:14 mithro: also - I don't know if it applies - encoding comes to mind as well Apr 15 14:26:23 forget that... Apr 15 14:26:31 don't worry about string conversion issues atm Apr 15 14:26:38 so do you actually have an implimentation? Apr 15 14:26:48 mithro: marc singer had one Apr 15 14:27:11 mithro: ftp.buici.com Apr 15 14:27:34 pub/oe/bbp-1.0.3.tar.gz Apr 15 14:28:03 mithro: We've never had the C<->Python data interchange - there was a implementation in C and obviously the python verison... Apr 15 14:28:20 RP: yeah i know i'm doing the C<->python data interchange :P Apr 15 14:28:44 mithro: :) Apr 15 14:33:13 zecke: so i just need to copy across mapped file? Apr 15 14:33:25 yes Apr 15 14:36:02 then how do i fix up input? Apr 15 14:36:45 lex_t has a file* make that a filedescriptor Apr 15 14:36:53 and copy the old code into input ;) Apr 15 14:37:17 * chouimat is back. Apr 15 14:37:31 morning Apr 15 14:40:04 hey chouimat Apr 15 14:40:18 ahh fsck Apr 15 14:40:27 now I have to call C++ code from C Apr 15 14:40:49 mithro: wrap it ;) Apr 15 14:56:26 03koen 07org.oe.dev * r3276dff2... 10/packages/freeciv/freeciv_2.0.8.bb: freeciv: add freeciv 2.0.8 Apr 15 15:04:46 actually Apr 15 15:04:49 zecke: you alive? Apr 15 15:10:19 hrm Apr 15 15:10:39 kind of Apr 15 15:13:04 do you know what a result from yylex of 1001 means? Apr 15 15:15:12 it appears to be T_EOF Apr 15 15:19:56 only none of the E functions are getting called Apr 15 15:21:57 not out of my head Apr 15 15:22:28 <> { return T_EOF; } Apr 15 15:22:46 this means it scanned to the EOF Apr 15 15:23:15 how can i find out how it got to EOF without doing anything Apr 15 15:23:16 if (result != T_EOF) Apr 15 15:23:16 printf ("premature end of file\n"); Apr 15 15:23:49 using bbparseTrace Apr 15 15:24:20 give it a FILE* Apr 15 15:30:57 i have Apr 15 15:32:53 mithro: for the tracing Apr 15 15:33:19 ahh Apr 15 15:35:34 i've found the problem Apr 15 15:35:47 e_parse_error Apr 15 15:35:47 Exception exceptions.NameError: 'CParseError' in 'bitbakec.e_parse_error' ignored Apr 15 15:37:23 know any way to get the parser error from yylex? Apr 15 15:38:09 mithro: see #define ERROR(e) in bitbakescanner.l Apr 15 15:41:55 zecke: Guess what's slowing down my cache the most - SkipPackage :) Apr 15 15:42:27 good Apr 15 15:42:49 good? Apr 15 15:45:18 RP: well, that is the bottleneck on cached loading now? Apr 15 15:45:42 zecke: Yes - as it reparses the skipped files each time :-/ Apr 15 15:46:09 do you have a very simple bitbake file with no inherits/etc Apr 15 15:46:31 every file inherits base.bbclass Apr 15 15:47:05 mithro: try a sample like bitbake.conf (and remove include... and name it .bb) Apr 15 15:47:17 okay Apr 15 15:47:27 is Apr 15 15:47:29 or a base.bbclass with lumps chopped out Apr 15 15:47:37 "a := a" a valid bitbake file? Apr 15 15:48:07 That's self referencing which I'm not sure we allow Apr 15 15:48:58 actually, its not as I'm thinking of a := ${a} :) Apr 15 15:49:57 RP: r412 also suffers from the cached-failure problem Apr 15 15:50:43 koen: ok, the issues I see for 420 onwards are a different problem then :-/ Apr 15 15:51:19 in this case it went straight to do_rootfs, which obviously failed with missing ~30 packages Apr 15 15:52:08 e_parse_error line: 2 parse: 1 Apr 15 15:53:10 koen: Its interesting my testing didn't show that up :-/ Apr 15 15:54:01 that seems to be errorUnexpectedInput Apr 15 15:55:15 so what is with "\na := a\n" Apr 15 15:57:51 anyone know if there has been any work done on the wgt634u router? Apr 15 16:01:28 not here AFAIK Apr 15 16:11:26 SkipPackage is a can of worms. I'd still like it to return the data dict even when it raises an exception :-/ Apr 15 16:14:25 Although maybe I'm speaking too soon - I perhaps just need to rework this... Apr 15 16:20:43 SkipPackage? Apr 15 16:24:10 morning Apr 15 16:25:37 morning tmm1 Apr 15 16:33:15 grr, I wonder why I'm getting this broken 2.6 kernel in my gpe-image build Apr 15 16:33:19 * pb_ stabs oe Apr 15 16:37:40 * chouimat is away: grocery Apr 15 16:41:32 zecke: so any idea why the parser is dieing on my test file? Apr 15 16:42:27 indigestion? Apr 15 16:43:01 the C one seems to like it okay Apr 15 16:44:28 Reparsing with a clean cache down to 3.3s :) Apr 15 16:44:45 RP: that's only 60 times faster Apr 15 16:44:56 go for 100 times faster ;) Apr 15 16:45:48 koen: It probably is 100 times faster. I'd be curious to see what ewi makes of it :) Apr 15 16:46:15 Initial parsing seems slower now though so I need to work out why... Apr 15 16:51:34 3.3 seconds?! is that allowed? Apr 15 16:52:20 RP: how does this effect the overall speed? Apr 15 16:53:01 mithro: In theory, it shouldn't change much at all. In practice, it might be slowing down the inital parse Apr 15 16:53:27 mithro: I'll finish off the code and run some proper before/after tests... Apr 15 16:53:28 if it wouldn't change much, why are you doing it? Apr 15 16:53:43 mithro: Ah, define what you mean by overall speed Apr 15 16:54:00 speed between bitbake xxx and it starting to bake it Apr 15 16:54:24 mithro: That's about 60+ times faster if nothing changed Apr 15 16:54:35 (i.e the cache is valid) Apr 15 16:54:39 but on the first run no change? Apr 15 16:54:55 mithro: no, I'm not working on that Apr 15 16:55:34 mithro: This should only reparse the files it needs to so if you change a single .bb and rerun, it will still have the speed gain. You only lose if you edit local.conf or similar Apr 15 16:56:36 I just touched the linux-openzaurus.inc file and I lost a whole extra 0.4 seconds in reparsing time :) Apr 15 16:56:46 RP: hehe, I hate that game Apr 15 16:56:57 * mithro pokes zecke Apr 15 16:57:02 i'm suck Apr 15 16:57:06 s/suck/stuck Apr 15 16:57:11 RP: HEAD started to build oz354fam083 without a issue Apr 15 16:57:35 i believe the problem is in the parser code Apr 15 16:58:01 zecke: I'm using .dev Apr 15 16:58:42 RP: I just wanted to test, as people report otherwise Apr 15 16:59:15 zecke: For reference, I use oz354x with revision 419 ok Apr 15 16:59:44 okay... Apr 15 16:59:46 It might throw odd parsing errors, I can't remember. That might be r423 Apr 15 17:00:03 RP: I throw one kind of error now Apr 15 17:00:14 RP: when two different files specify a method with the same name... Apr 15 17:01:16 Exception exceptions.AttributeError: "'list' object has no attribute 'setVar'" in 'bitbakec.e_immediate' ignored Apr 15 17:01:18 :P Apr 15 17:02:00 zecke: This is with head? Have you committed anything recently? Apr 15 17:05:48 In case anyone fancies experimenting, this patch adds what I've done so far and reverts the changes between r419 and r423: http://www.rpsys.net/openzaurus/temp/bitbake_newcache-r0.patch Apr 15 17:06:02 Yes, I know writing to /tmp is bad and I will fix that ;-) Apr 15 17:06:16 I'm going to do something else for a bit... Apr 15 17:08:21 mithro: you pass the wrong Pyobject? Apr 15 17:08:44 zecke: yeah delebrately Apr 15 17:09:02 Hi again guys Apr 15 17:09:03 *depressed* Apr 15 17:09:10 hey Ecco Apr 15 17:09:15 ;-) Apr 15 17:09:29 zecke: want some vla? Apr 15 17:09:49 koen: want a Diploma... but not the ones from the SPAM mails Apr 15 17:10:06 I have 11 exams to take... I'm collecting them... Apr 15 17:10:11 ouch Apr 15 17:10:39 03koen 07org.oe.dev * rfbe3dd34... 10/ (3 files in 3 dirs): Apr 15 17:10:39 efika.conf, linux-efika: add files to get rudimentary support for the efika machine into OE. Apr 15 17:10:39 * see http://www.pegasosppc.com/efika.php for more details Apr 15 17:10:50 koen: non is actually challenging... I take the written ones, 'oral' Apr 15 17:10:58 koen: but I have to make the dates myself Apr 15 17:10:59 I was wondering something about xtscal : how does it work (from the user point of view) ? Apr 15 17:11:12 Ecco: you click at five positions Apr 15 17:11:22 Ecco: using something like a pseudo pencil Apr 15 17:11:38 ok Apr 15 17:11:49 because I'm having some troubles with my current distro Apr 15 17:12:13 may i do something like "Xfbdev & export DISPLAY=:0 ; xtscal" Apr 15 17:12:20 ? Apr 15 17:12:25 because that doesn't display anything Apr 15 17:12:36 (apart from the base Xfbdev of course) Apr 15 17:12:36 Ecco: should work Apr 15 17:12:40 ok Apr 15 17:12:43 Xfbdev -ac & Apr 15 17:12:45 Ecco: you can also delete the calibration file Apr 15 17:12:48 where am I supposed to press the screen ? Apr 15 17:13:29 koen, : what does ac do ? It says "disable access restriction", but what does that mean ? Apr 15 17:13:56 it means you can connect from outside an xauth session Apr 15 17:14:09 ha. Which means ? Apr 15 17:14:16 (sorry, not really an X guru) Apr 15 17:15:43 it means that if you have trouble connecting to the X server you might have better luck with the access controls turned off Apr 15 17:16:16 oh, ok, that's a neat explaination :-) Apr 15 17:17:59 ahem Apr 15 17:18:22 well, indeed Xfbdev -ac & export DISPLAY=:0 ; xtscal does not work Apr 15 17:18:43 it does nothing more than raw "Xfbdev" Apr 15 17:19:01 any errors on the console? Apr 15 17:19:20 nope :( Apr 15 17:19:27 and logread doesn't give anything either Apr 15 17:19:47 ah Apr 15 17:19:53 yeah, it says "Cannot open XFT font" Apr 15 17:19:59 May it be that ? Apr 15 17:20:14 Ecco: that is totally Off topic here Apr 15 17:20:23 zecke, yeah, I know. Sorry :-$ Apr 15 17:20:35 Ecco: try to find another channel Apr 15 17:20:41 ok Apr 15 17:20:43 e.g. #gpe or #freedesktop.org Apr 15 17:20:49 thx :-) Apr 15 17:21:02 or even #openzaurus-users (or whatever it is named) Apr 15 17:26:57 has anyone seen this error trying to build tcl: checking system version (for dynamic loading)... /opt/oe/build/tmp/work/arm-linux/tcl-8.4.11-r2/tcl8.4.11/unix/configure: line 7624: syntax error near unexpected token `)' Apr 15 17:27:34 mreimer_: not me, but I can't remember to have build tcl Apr 15 17:27:52 python is trying to build it for some reason Apr 15 17:29:09 mreimer_: you can safely remove tcl (and tk) from pythons depends Apr 15 17:29:52 thanks koen Apr 15 17:30:45 koen, finally got abiword-plugins built. thanks again. I hope to try out the OOo plugin soon Apr 15 17:31:13 mreimer_: how does the line look like? Apr 15 17:31:13 koen: why? Apr 15 17:31:58 zecke: it's optional for python and doesn't build in some configs Apr 15 17:32:14 I couldn't find a problem in the configure itself Apr 15 17:32:56 zecke: OSF*) Apr 15 17:33:10 I looked in the surrounding context, but didn't see any obvious syntax errors. Apr 15 17:33:31 I'll go blind if I spend too much time trying to debug a configure script Apr 15 17:33:41 maybe it's a shell problem Apr 15 17:34:58 or some badly written case statement Apr 15 17:35:12 looks fine to me Apr 15 17:35:16 thanks for the help guys. gotta run Apr 15 17:49:26 hey pH5 Apr 15 17:49:56 yay its working! Apr 15 17:51:34 mithro: the c/c++ based parser? Apr 15 17:52:20 yeah Apr 15 17:53:35 is it fast? Apr 15 17:53:50 hey koen, good evening everyone Apr 15 17:56:01 dunno Apr 15 17:56:33 in theory it should be Apr 15 17:56:37 pH5: do you think you can find some time to port the linux-oz eabi selectio ot hh-pxa? Apr 15 17:56:44 but a lot of the speed is lost in other places Apr 15 18:01:15 * pH5 pulls .dev Apr 15 18:02:27 pH5: I pushed a new gtk-engines which should fix the abiword crash with clearlooks Apr 15 18:03:23 anyone know how to use svk? Apr 15 18:03:39 mithro: there's a short howto in the oe wiki Apr 15 18:04:05 mithro: http://oe.handhelds.org/cgi-bin/moin.cgi/SvnScmTrial Apr 15 18:10:12 SVK not svn? Apr 15 18:10:26 mithro, read the whole page? Apr 15 18:10:31 yeah Apr 15 18:10:34 i just realised that Apr 15 18:10:54 There's a bunch of svk commands Apr 15 18:11:00 (not that I read them:)) Apr 15 18:17:52 re Apr 15 18:18:54 Bernardo: wb Apr 15 18:19:55 * Bernardo 's been trying to make some sense out of his expenses and earnings Apr 15 18:20:00 koen: yay, freeciv :) Apr 15 18:20:09 :) Apr 15 18:20:15 03mwester 07org.oe.dev * ref08aba7... 10/ (12 files in 6 dirs): Unslung: unslung-rootfs, unslung-kernel, unslung-image - rearranged NLS support, removed Linksys upgrade page Apr 15 18:21:01 koen: I'm dense. where do I find the linux-oz eabi selection? Apr 15 18:21:30 pH5: packages/linux/linux-openzaurus.inc Apr 15 18:21:46 koen: heh, .inc. thanks Apr 15 18:22:04 phew Apr 15 18:22:11 has anyone played with mDNS and zeroconf? Apr 15 18:22:55 I toyed briefly with avahi Apr 15 18:23:10 but pH5 is the avahi maestro in here Apr 15 18:24:55 just reading up on it, pretty cool stuff. especially for usb gadget devices. i found an ARP broadcast kernel patch, a patch for udhcpc that reverts to zeroconf is no dhcp server is found, and a bunch of tiny userspace daemons Apr 15 18:28:08 dhcp is so ipv4 Apr 15 18:28:59 heh Apr 15 18:29:16 PXEboot isn't yet ipv6 though Apr 15 18:29:22 is it? Apr 15 18:32:35 no idea Apr 15 18:32:41 I use nfs or tftp Apr 15 18:33:00 ~stab redboot Apr 15 18:33:02 * ibot runs at redboot with an origami Swiss Army knife, and inflicts a nasty paper cut. Apr 15 18:37:08 pH5: great news about hx4700 wlan, isn't it? Apr 15 18:41:59 koen: yeah, especially as I don't even have a wlan cf card. Apr 15 18:46:11 is your microdrive still broken? Apr 15 18:46:18 * koen is thinking about getting one Apr 15 18:49:19 koen: isn't this hx device discontinued? Apr 15 18:49:41 zecke: iirc not, but I was talking about the MD Apr 15 18:52:41 koen: I sent it in, I hope the replacement will arrive next week. Apr 15 18:52:41 koen: http://en.pastebin.ca/49436 Apr 15 18:54:16 pH5: looks ok to me Apr 15 18:55:40 * RP returns Apr 15 18:56:35 hey zecke, RP Apr 15 18:56:52 RP: wb Apr 15 19:01:33 ooooooooh Apr 15 19:01:34 http://linuxdevices.com/news/NS6404482880.html Apr 15 19:01:41 5x gbit Apr 15 19:05:52 hmh, nice. Apr 15 19:07:56 6x gbit Apr 15 19:08:38 even better! Apr 15 19:09:12 grr usb gadget ether Apr 15 19:09:27 hmmm... might turn out to be 4x gbit only (if the gbit port isn't somehow connected to the gigabit switch) Apr 15 19:09:28 ah, I see Apr 15 19:09:42 6x phy + 5x switch + 1 "simpleport" Apr 15 19:10:14 it does seem to feature 6x rj45 Apr 15 19:10:44 koen, yep, 3x2 Apr 15 19:11:19 if you look at the large picture you can see 6x YCLPG243002 Apr 15 19:11:39 and one big and one small vitesse chip Apr 15 19:11:41 yup Apr 15 19:12:10 "along with a standards-based Linux software stack " Apr 15 19:13:18 4x sata, 1x pata, 4x usb, 1x pci, 1x (168pin) sdram.. nice Apr 15 19:15:01 apple switching to intel really got freescale into action it seems Apr 15 19:16:38 if you want to have free powerpc hardware: http://projects.ppczone.org/projects.php Apr 15 19:17:14 evildevil: http://www.freescale.com/webapp/sps/site/prod_summary.jsp?code=MPC8349E&nodeId=018rH3bTdG8657 Apr 15 19:17:25 evildevil: it only has 2 gbit port on-chip Apr 15 19:17:36 ports* Apr 15 19:18:32 koen, do you know how much this board is going to cost? Apr 15 19:18:38 I have no idea Apr 15 19:18:48 I can't seem to find it on the freescale site Apr 15 19:18:59 but I suspect $300 - $400 Apr 15 19:22:44 yeah, this seems to be a reasonable price. Apr 15 19:24:08 this thing could replace my switch, my nslu2, my external usb-hdd (including 3 power supplies) Apr 15 19:24:39 that's what I was thinking as well Apr 15 19:26:33 it seems to be powered by a standard atx power supply (there must be some power connectors for the internal hdd's somewhere..) Apr 15 19:41:20 didn't sata have some funky powerconnectors? Apr 15 19:41:25 * koen is still PATA only Apr 15 19:53:35 Current sata have both and old style and new style connector Apr 15 19:53:47 zecke: Are you busy? Apr 15 19:55:35 RP: hehe Apr 15 19:55:48 RP: I played with addpart and my backup disk is busted :} Apr 15 19:56:07 zecke: Oops :-/ Apr 15 19:56:21 zecke: I'm pondering what to do with the original bitbake cache Apr 15 19:56:30 RP: I gave a wrong length on addpart and created a fs :} Apr 15 19:56:50 RP: throw it away ;) Apr 15 19:57:02 zecke: I guess you'll not make that mistake again ;-) Apr 15 19:57:16 RP: we could replace it with something a Cache with ~30 slots Apr 15 19:57:30 zecke: I'm just not sure how much of the data store I have to remove/fix... Apr 15 19:57:31 RP: did you make that mistake as well (once) Apr 15 19:57:46 zecke: I've made similar mistakes ;-) Apr 15 19:58:11 RP: I'm currently copying the other parts and playing somemore with delpart and addpart ;) Apr 15 19:58:32 zecke: You evidently like taking risks :) Apr 15 19:58:34 RP: ah and GNU parted doesn't want to change my part table Apr 15 19:58:49 Use fdisk ;-) Apr 15 19:59:12 RP: how can I change existing partitions? I would need to del and create one Apr 15 19:59:17 okay back to the cache Apr 15 19:59:27 RP: what is your plan? Apr 15 19:59:29 zecke: Ah, fdisk doesn't do that :) Apr 15 19:59:49 zecke: My current code basically removes the need for the current form of caching Apr 15 20:00:05 The current cache causes us 400MB of IO for no gain I can see now... Apr 15 20:00:17 I've proven my replacement can reparse in seconds Apr 15 20:00:26 RP: do you cache enough data to do findBestProvider? Apr 15 20:00:35 zecke: Easily Apr 15 20:01:01 zecke: Its also dynamic - if findBestProvider uses a new variable, the cache adapts :) Apr 15 20:01:05 hey gerwinin Apr 15 20:01:19 zecke: Did you look at the patch? Apr 15 20:01:30 RP: have you posted that somewhere? Apr 15 20:01:49 zecke http://www.rpsys.net/openzaurus/temp/bitbake_newcache-r0.patch Apr 15 20:01:55 zecke: I mentioned it earlier Apr 15 20:02:33 RP: oh Apr 15 20:02:33 (Ignore the bits that revert r423 to r419 - they're two isolated files atm Apr 15 20:02:39 - if not var in self.dict: Apr 15 20:02:40 - self._makeShadowCopy(var) Apr 15 20:02:44 that was new as well Apr 15 20:03:08 zecke: This is a working copy diff that also reverts r423 to r419 Apr 15 20:03:33 I don't actually need to touch data.py, data_smart.py (yet) Apr 15 20:03:53 Ignore changes to those two files Apr 15 20:04:03 RP: feel free to commit fetch/__init__.py Apr 15 20:04:14 zecke: I still need to commit it all Apr 15 20:04:35 koen|tv, the only sata disk i own does have a special sata power connector, but it came with an adapter Apr 15 20:04:46 zecke: If I remove the orginal cache, it gets much simpler though so no point in commenting until we decide which way to go Apr 15 20:04:53 RP: rppickle, worse than the BSD advertising clause Apr 15 20:05:03 zecke: :) Apr 15 20:05:14 zecke: That will be fixed to use the cache directory ;-) Apr 15 20:05:51 he he Apr 15 20:05:53 Obvious Americans twice as likely to view Islam favorably than Scientology. Tom Cruise envious of Bin Laden's recruiting ability Apr 15 20:07:11 RP: how do we stop the cache growing to hunderds of MB? Apr 15 20:07:32 zecke: By being careful about the number of variables you access via bb_cache Apr 15 20:08:42 RP: let us play bitbake now Apr 15 20:08:46 RP: bitbake opie-image Apr 15 20:09:18 RP: we build the dependencies, will we reread the bbfile for quilt-native more than once? Apr 15 20:10:01 RP: we will have all information cached to not load the file a second time and find out that we have it already build? Apr 15 20:10:04 When you build the dependencies, no, just once. It will be read once more when quilt-native builds Apr 15 20:10:30 But I'm arguing its faster to reparse the single .bb files as we build them Apr 15 20:10:44 RP: I think we can remove the current caching Apr 15 20:10:49 (than spend 400MB of disk IO caching them all just in case) Apr 15 20:11:07 zecke: Can you see how this is working? Apr 15 20:12:55 RP: I like the approach Apr 15 20:13:08 RP: I still have to find out about getVar ;) Apr 15 20:13:56 zecke: getVar watches which variables are accessed through bb_cache. It only saves the ones that are accessed Apr 15 20:14:18 This means our total cache file is 8MB large Apr 15 20:14:39 that doesn't fit on my 5000d, so it sucks Apr 15 20:14:44 *cough* Apr 15 20:16:41 RP: remove the old caching Apr 15 20:16:53 RP: what about SkipPackage? Apr 15 20:17:12 zecke: I have that handled happily now Apr 15 20:17:39 We save a note it was skipped (and its __depends) Apr 15 20:17:41 ah okay within the cache Apr 15 20:17:48 zecke: Does this have any implication for DataSmartPackage? Apr 15 20:18:38 RP: we can remove some code Apr 15 20:18:53 but I'm glad to remove it Apr 15 20:19:16 zecke: ok. I'm not 100% what it does but I'll have a look :) Apr 15 20:19:26 zecke: The only casualty is that after these changes is bb shell's poke doesn't work Apr 15 20:19:52 zecke: I think its a small price to pay and I suspect mickeyl can find a way around it :} Apr 15 20:20:17 RP: well, we can move the 'dirty' handling to the Cache Apr 15 20:20:39 RP: loadFullData would do this then?! Apr 15 20:20:41 zecke: The problem isn't dirty handling - pkgdata doesn't exist anymore Apr 15 20:21:56 RP: the only thing I need to point out Apr 15 20:22:03 RP: Document the Cache Apr 15 20:22:14 zecke: I promise to :) Apr 15 20:22:15 RP: the idea under class Cache:\n"""Here is my plan...""" Apr 15 20:22:39 zecke: If we rip out the old cache it gets much simpler and easier to document Apr 15 20:22:49 RP: rip it out Apr 15 20:22:58 zecke: ok :) Apr 15 20:23:00 RP: we might have the C parser done soon as well thanks to mithro Apr 15 20:23:15 RP: it gives so much issues... Apr 15 20:23:23 RP: One more question Apr 15 20:23:24 zecke: As they'll speed up different things, this would be great :) Apr 15 20:23:39 RP: do you cache the variables expanded or unexpanded? Apr 15 20:23:44 Which gives issues - the C parser or the cache? Apr 15 20:23:48 zecke: Expanded Apr 15 20:24:00 RP: that needs to be changed Apr 15 20:24:08 zecke: Why? Apr 15 20:24:09 RP: think of the moving SRCDATE Apr 15 20:25:41 zecke: If you want to do that, the whole idea is dead in the water Apr 15 20:26:06 why? Apr 15 20:26:58 You can't use the unexpanded versions in the likes of findBestProvider and we don't cache enough to expand them Apr 15 20:27:38 expanding them also means potential calls to python methods and is way out of the scope of what the cache code currently does Apr 15 20:29:31 (Its also partially broken in the existing bitbake caching) Apr 15 20:29:50 right Apr 15 20:30:02 RP: okay this means we need to reparse once a day? Apr 15 20:30:29 zecke: For the change to DATE, yes Apr 15 20:31:05 For TIME, once per second :} Apr 15 20:31:34 {: Apr 15 20:34:53 RP: hmm not too good then Apr 15 20:35:27 ah delpart and addpart worked Apr 15 20:35:36 I just had to redo my error ;) Apr 15 20:36:26 zecke: It will only use the cached data to make the decisions about which packages to build. It will reexpand the variables when it acutally builds a provider Apr 15 20:36:46 I realise this could get slightly confusing :-/ Apr 15 20:37:46 RP: just call it 'MagicCache' Apr 15 20:38:29 To be honest, I think people would prefer the speed than 100% cache accuracy. If they want accuracy, they can reparse... Apr 15 20:38:49 If only reparse reparsed! :) Apr 15 20:39:47 RP: hmm I wanted to say I'm offline for an hour Apr 15 20:40:01 RP: but I decided to be offline tomorrow and monday so I can hack now Apr 15 20:40:55 zecke: Don't let me put you off exam revision or whatever - this can wait Apr 15 20:41:18 RP: it is Wednesay and Friday Apr 15 20:41:28 RP: and I think I already know enough to pass it Apr 15 20:42:20 I used to get addicted to new TV series and spider solitaire before exams :-/. Anything but revision :) Apr 15 20:42:32 heh Apr 15 20:42:35 spider solitaire Apr 15 20:42:59 and watching stargate sg1's first 6 seasons in about 2 weeks Apr 15 20:43:16 koen: invite me the next time Apr 15 20:44:12 koen: does a 'diff' ipk script exist? Apr 15 20:44:56 zecke: How about I put it this way - Our current cache isn't scalable. My proposal is more scalable and the drawback you mention isn't a fundamental part of the design Apr 15 20:45:04 zecke: just unpack the ipkgs? Apr 15 20:45:30 There is nothing stopping us changing it to use unexpanded data, if we can work the depencencies out Apr 15 20:45:46 right Apr 15 20:46:05 (Which may be possible which a cache dummy front to the data store to see which variables we look at) Apr 15 20:46:31 I'll adapt the patch to remove the existing cache and see where I end up Apr 15 20:54:47 right Apr 15 20:55:38 oh well, I watch that movie and then I'm going to read some more Apr 15 21:05:53 RP: Did you get that sound interrupt rewrite done that you mentioned to me before? Apr 15 21:06:10 JustinP: Sound interrupt? Apr 15 21:06:55 I shall quote from....months ago Apr 15 21:07:04 +12:13 < RP> JustinP: There is going to be an issue with it fighting the sound system for the headphone interrupt Apr 15 21:07:07 +12:14 < RP> JustinP: I'm about to rewrite the headphone interrupt handling anyway so the best soution might be to just comme\ Apr 15 21:07:10 nt it out in the sound code Apr 15 21:07:17 sorry, copied from my patch :-/ Apr 15 21:07:30 Yes, that's all handled by the input system now Apr 15 21:07:41 and even in mainline :) Apr 15 21:07:44 cool Apr 15 21:07:55 so I'm going to need to fix the remote then Apr 15 21:08:16 in fact, the patch probably doesn't fully apply anymore anyway....I'm about to try to get it to apply to 2.6.16 Apr 15 21:08:54 Instead of disabling it in the sound driver, you need to disable in the keyboard driver (or better, share it) Apr 15 21:09:04 drivers/input/keyboard/spitkbd.c Apr 15 21:09:04 once I know it applies could you take a look at the interrupt and let me know what needs to be changed? Apr 15 21:09:16 To disable it? Apr 15 21:09:24 disabling the keyboard interrupt doesn't sound like a good idea to me.... Apr 15 21:09:39 No, just the headphone interrupt Apr 15 21:09:51 There are lots of interrupts in the keyboard driver... Apr 15 21:10:04 ok Apr 15 21:10:16 well, sharing also sounds like a good idea.... Apr 15 21:10:49 It would be best for you to talk to someone who's shared interruts before - I haven't so I don't know what the pitfalls are Apr 15 21:11:03 hmmmm...fun.... Apr 15 21:11:27 RP: oh Apr 15 21:11:32 To disable, basically find the line that says request_irq*AK_INT and comment it out in the file I mentioned Apr 15 21:11:35 so the headphone interrupt handles insert/eject of headphone jack, right? Apr 15 21:11:40 yes Apr 15 21:11:54 ok Apr 15 21:12:02 zecke: oh? Apr 15 21:12:03 well, I'll see if I can get this thing working again Apr 15 21:15:07 RP: good stuff Apr 15 21:15:13 mickeyl: hey Apr 15 21:15:20 hi mickeyl :) Apr 15 21:15:31 I will shed a tear when you rip off the cache though :)) Apr 15 21:15:47 RP: could you try r433 and see if that changed anything Apr 15 21:16:05 mickeyl: Its really the same code with a slightly different method ;-) Apr 15 21:16:09 RP: if it did, I will explain the oh Apr 15 21:16:20 mickeyl: another pattern is applied ;) Apr 15 21:16:21 mickeyl: I'm afraid I'll break the shell's poke command but I there are ways it can be fixed again Apr 15 21:16:34 RP: I broke it as well ;) Apr 15 21:16:36 RP: right. Apr 15 21:16:40 RP: don't worry about that Apr 15 21:18:30 times have changed. when i added the cache, we had less metadata and parsing was even more infefficient. Apr 15 21:18:34 good to have a new scheme now Apr 15 21:18:43 hey mickeyl Apr 15 21:18:46 Guys Apr 15 21:18:58 Is it ok to report some Zaurus kernel bugs into OE's bugtracker ? Apr 15 21:19:04 mickeyl: The fast reparse is quite scary when you're used to the old bitbake :) Apr 15 21:19:15 RP: indeed. looks like it's just doing nothing Apr 15 21:19:33 Ecco: yeah, go ahead. use oz 3.5.4.x as product Apr 15 21:19:48 Ok Apr 15 21:19:53 I think that'll interest RP Apr 15 21:20:07 Remember that USB_gadget bug ? Apr 15 21:20:23 Ecco: yes Apr 15 21:20:25 RP: I guess with that few data to cache it makes no sense to try to use marshall instead of pickle. we wouldn't gain much now Apr 15 21:20:38 It looks like it is happens because USB_host is hardcoded within OZ 3.5.4.1 Akitta's kernet Apr 15 21:20:40 kernel Apr 15 21:20:43 does that make sense ? Apr 15 21:21:11 Ecco: Kind of. Do you know where its hardcoded? Apr 15 21:21:24 What do you mean ? Apr 15 21:21:25 mickeyl: Its a single pickle/unpick operation now on an ~8MB file :) Apr 15 21:21:33 I mean that actually usbcore is not a module Apr 15 21:21:36 what is the usb gadget bug? Apr 15 21:21:39 it's built straight into the kernel Apr 15 21:21:49 Ecco: And that makes the difference? Apr 15 21:21:51 TheMasterMind1, on my C-1000 usb gadget acts really weird Apr 15 21:21:57 RP : I don't know wether that's it Apr 15 21:22:02 I rolled out my own kernel Apr 15 21:22:07 without usbcore at all Apr 15 21:22:09 i'm trying to get usb gadget ether recognized by osx Apr 15 21:22:14 Ecco: usbcore shouldn't make any difference at all if ohci_hcd isn't loaded Apr 15 21:22:32 TheMasterMind1 : g_ether works great with OSX out of the box Apr 15 21:22:33 ohci_hcd is what talks to the hardware, nothing else. Apr 15 21:22:37 hmmm Apr 15 21:22:43 Ok, so it might be not that Apr 15 21:22:49 But it's really "random" Apr 15 21:22:57 I thought it could be related to Power Managment Apr 15 21:23:10 If you can show that ohci_hcd leaves the machine in the wrong state to support UDC, that would be a bug Apr 15 21:23:11 but it looks like it works (and doesn't) both with AC or not Apr 15 21:23:24 hmm, indeed I never used ohci_hcd Apr 15 21:23:49 but ohci_hcd was autoloaded by your original OZ supplied kernel Apr 15 21:23:50 Ecco: strangely enough it isn't for me. osx says initDevice failed Apr 15 21:24:03 TheMasterMind1, Does OSX enumerate your device ? Apr 15 21:24:12 In fact the problem is not with g_ether Apr 15 21:24:14 RP: now that the command line is so fast, one of the reasons for the shell vanished. I need to improve it to drag people into using it :D Apr 15 21:24:31 it's with the pxa27x__udc Apr 15 21:24:39 * polyonymous is using shell Apr 15 21:24:47 ! :) Apr 15 21:24:48 mickeyl: I have no frelling idea! Apr 15 21:24:51 mickeyl, speaking of which. Apr 15 21:24:53 mickeyl: I take bribes in the form of Leffe Radieusse Apr 15 21:25:03 koen: hehe. Apr 15 21:25:21 mickeyl, when I say 'reparse' I want it to reparse even if it thinks the file is up to date, but I know included files aren't :) Apr 15 21:25:24 Ecco: enumerate? Apr 15 21:25:28 Yeah Apr 15 21:25:38 mickeyl: I can't wait to see what these impovements will be :) Apr 15 21:25:41 polyonymous: right. that timestamp check is bogus. I'll remove it asap Apr 15 21:25:50 Basically the host PC doesn't even see the device (the Zaurus) Apr 15 21:25:58 I'm still dreaming about complete GUI frontend... Apr 15 21:26:05 mickeyl, just wanted to make sure you know about it. Apr 15 21:26:09 mickeyl, _G_UI? Apr 15 21:26:09 mickeyl: and 'addparse' for new files/versions Apr 15 21:26:11 polyonymous: yeah Apr 15 21:26:17 mickeyl: Once I write this multithreaded bitbake, it'll need a UI and I'm hoping to talk to you about it when it gets to that point ;-) Apr 15 21:26:21 Why would one want the 'G' part there?? Apr 15 21:26:21 yes, GUI as in graphical :)) Apr 15 21:26:29 *G*ui, not a *Q*ui Apr 15 21:26:31 ;) Apr 15 21:26:34 Oui :) Apr 15 21:26:35 Ecco: well in this case i'm working with a pxa255 based gumstix Apr 15 21:26:36 heh Apr 15 21:26:44 RP: sounds good Apr 15 21:26:55 Well, I can't imagine the special graphical effects you have to offer :) Apr 15 21:27:22 mickeyl: I think the timestamp check might get fixed in my current rearrangement of things Apr 15 21:27:23 TheMasterMind1, ok, so if you do "Apple -> About this Mac -> More Info -> USB", can you see the gumstick ? Apr 15 21:27:28 JustinP: ping Apr 15 21:27:34 hey XorA|gone Apr 15 21:27:35 kind of browsing would be nice though, but I was thinking about making a web frontend for it. Apr 15 21:27:50 polyonymous: it would be more than just a GUI for what bitbake can do now. I'm thinking more along the lines of a Bitbake IDE. like, adding a new distro wizard, adding a new .bb, debugging .bb's etc. Apr 15 21:27:55 RP: ah, yeah Apr 15 21:27:57 hey koen Apr 15 21:28:00 morning Apr 15 21:28:01 Ecco: ah didn't even know that was there. indeed i see a 'RNDIS/Ethernet Gadget' listed Apr 15 21:28:06 mickeyl, I see. That's different. Apr 15 21:28:34 mickeyl: eclipse integration Apr 15 21:28:34 TheMasterMind1, so you're good with that Apr 15 21:28:41 did I really say that? Apr 15 21:28:43 So far I'm happy with the shell -- saves time tremendously. Apr 15 21:28:45 Now it's easy : modprobe g_ether on the device Apr 15 21:28:48 * koen washes mouth with soap Apr 15 21:28:52 bring usb0 up with an ip addres Apr 15 21:28:54 java zealot :D Apr 15 21:28:58 get to "System preferences, Network" Apr 15 21:29:05 and it says "A new device appeared" Apr 15 21:29:07 use a leffe, not soap Apr 15 21:29:09 and you' ready to go Apr 15 21:29:33 XorA: Pong? Apr 15 21:29:45 JustinP: answered your email about gnome-vfs-dbus Apr 15 21:30:26 Ecco: nope, no go. if i bring up system.log i can see a 'AppleUSBCDC: start - initDevice failed' Apr 15 21:30:33 Hmm Apr 15 21:30:35 that's weird Apr 15 21:30:41 Maybe you're using an old kernel ? Apr 15 21:30:48 I've been using it on OMAP and on PXA Apr 15 21:30:51 XorA: I don't see anything.... Apr 15 21:30:53 zecke: My parsing errors don't show under r433 Apr 15 21:30:58 as long as the udc worked, g_ether worked well :-) Apr 15 21:31:00 i'm using 2.6.11 here Apr 15 21:31:08 (apart from the duplicate methods ones) Apr 15 21:31:24 ok so problem is in udc, that helps Apr 15 21:31:37 enable_irq(58) unbalanced from bf0514bc on module load Apr 15 21:31:39 TheMasterMind1, it may not be in udc Apr 15 21:31:48 I think it's not since your device is enumerated properly Apr 15 21:31:54 JustinP: literally just answered, let the mail servers catch up :-) Apr 15 21:31:57 Try to switch to a recent version Apr 15 21:32:00 RP: but the dependency issue is still present? Apr 15 21:32:01 like 2.6.16 or so Apr 15 21:32:19 alright, will try Apr 15 21:32:25 RP: I applied a hunk of your patch Apr 15 21:32:29 zecke: No, its not showing it Apr 15 21:32:31 * koen points to http://bugs.openembedded.org/show_bug.cgi?id=812 Apr 15 21:32:43 zecke: Which hunk? Apr 15 21:32:50 - if not var in self.dict: Apr 15 21:32:50 - self._makeShadowCopy(var) Apr 15 21:33:11 in the train I thought, we need to shadow it first Apr 15 21:33:22 to not influence the other users Apr 15 21:33:28 zecke: So I was sort of right :) Apr 15 21:33:40 RP: at least you forced me into a review Apr 15 21:34:32 zecke: My current bitbake is messed up and those test results are invalid, sorry :-( Apr 15 21:34:39 ~lart svn Apr 15 21:34:40 * ibot wallops svn with a main rotation server that needs rehubbing. It won't take long Apr 15 21:34:43 oh Apr 15 21:35:01 anyway, the lines were wrong, or simply not needed Apr 15 21:36:51 hmm with 3 seconds we should probably tweak or remove the progress bar altogether Apr 15 21:37:18 mickeyl: We still need it for the inital parse Apr 15 21:37:27 oh right Apr 15 21:37:37 * mickeyl tends to forget about that Apr 15 21:37:54 mickeyl: and it is nice to watch spinning like crazy on the reparse :) (I tried removing it and it didn't seem to make much difference) Apr 15 21:38:13 RP, try it over 2400 line :) Apr 15 21:38:19 RP: yeah that was my finding as well Apr 15 21:38:54 polyonymous: pipe to a file ;-) Apr 15 21:39:11 RP, piping progess meter to a file is the most brilliant idea ever! :) Apr 15 21:40:19 zecke: Having updated to your latest changes properly, the parser is flying :) Apr 15 21:40:41 RP: hmm it still fails on matchbox Apr 15 21:42:35 zecke: A full parse was 7 minutes before, now without the cache files to write and with your changes, its 2.45 mins :) Apr 15 21:42:45 zecke: and the quilt dependency issue didn't show up Apr 15 21:42:59 RP: oh well, I have the same issue I had before :} Apr 15 21:43:08 RP: but that is due cached loading it seems Apr 15 21:43:11 just wait for mithro to say "20 seconds" :) Apr 15 21:43:34 koen: My cache will still be faster ;-) Apr 15 21:43:34 RP: push it, push it, push it Apr 15 21:43:41 says my evil mind Apr 15 21:44:14 03koen 07org.oe.dev * r4ed4d06c... 10/packages/binutils/ (11 files): binutils_*.bb: apply patch from OE bug #812 to all binutils versions Apr 15 21:46:22 ok, latest patch: http://www.rpsys.net/openzaurus/temp/bitbake_newcache-r2.patch Apr 15 21:46:39 Applies to head, doesn't revert anything hopefully. Apr 15 21:46:56 I need to filter out the fetcher change that's lost in there but it won't hurt anyone Apr 15 21:47:08 RP: do we want to sync on keyboard interrupt? Apr 15 21:47:15 zecke: yes Apr 15 21:47:27 zecke: It will then continue parsing where it left off Apr 15 21:47:28 def delVar(self,var): Apr 15 21:47:28 + if not var in self.dict: Apr 15 21:47:29 + self._makeShadowCopy(var) Apr 15 21:47:32 it reverts that part Apr 15 21:47:35 :( Apr 15 21:48:09 RP: make sure to not commit fetch/__init__.py accidently Apr 15 21:48:16 RP: feel free to commit it though Apr 15 21:48:51 zecke: What do you think about iterating over the src mirror like that? Apr 15 21:48:59 RP: I'm fine with that Apr 15 21:49:13 I'll clean up and commit it separately Apr 15 21:50:22 RP: We can remove DataSmartPackage completely Apr 15 21:50:27 RP: wait a second Apr 15 21:50:43 zecke: I thought that might be possible but was going to ask you Apr 15 21:50:47 ah no Apr 15 21:51:23 oh welll ;) Apr 15 21:51:29 we can only use createCopy Apr 15 21:51:46 we won't have a 'detached' Dict Apr 15 21:54:05 RP: I'm getting tons of kernel-abiversion errors Apr 15 21:54:07 weird Apr 15 21:54:28 zecke: I was seeing them earlier I think Apr 15 21:54:46 RP: it is looking tim TMPDIR, so it can't find them... Apr 15 21:55:29 XorA: speeling cheecker ;) Apr 15 21:55:29 now a way to 'diff' the results of a bitbake run would be nice Apr 15 21:56:21 zecke: Ah, those variables - they've caused me problems in the past as you can only evaluate them properly after the kernel is staged :-/ Apr 15 21:56:39 RP: hmm, building quilt fails Apr 15 21:56:46 RP: smells like an update_data issue Apr 15 21:56:56 zecke: With the error I had? Apr 15 21:57:30 RP: okay with the cache it takes about 3-4 seconds Apr 15 21:57:46 RP: But I'm getting tons of RDEPENDS failures here Apr 15 21:58:22 zecke: Can you pastebin it? Apr 15 21:59:01 RP: ah... Apr 15 21:59:18 Rhttp://oe.pastebin.com/662186 Apr 15 21:59:30 every kernel module failed with RDEPENDS Apr 15 21:59:43 because there were parse errors with them before Apr 15 22:00:31 zecke: Ah, if the parsing aborts, that might leave the cache in a bad state. We should really trap that and delete the cache entry upon any parse error Apr 15 22:01:13 I just had a disturbin realization about the bitbake C parser Apr 15 22:01:15 "${@base_read_file('${STAGING_KERNEL_DIR}/kernel-abiversion')} Apr 15 22:01:23 koen: yes? Apr 15 22:01:32 zecke: I've always thought this was evil... Apr 15 22:01:33 it will probably break with -OMG -gentoo cflags Apr 15 22:01:56 mithro: can people pass CFLAGS with pyrex? Apr 15 22:02:19 koen: we probably run unittests on loading the module Apr 15 22:02:29 koen: and if it fails we fall back to python Apr 15 22:02:33 heh Apr 15 22:02:39 anyway Apr 15 22:02:42 'night all Apr 15 22:02:46 zecke: Those RDEPENDS certainly look like the original parsing failed corrupting the cache Apr 15 22:02:51 'night koen|sleep Apr 15 22:02:55 koen: or we put #ifdef GNUC > 2 #error "Unsupported compiler" #endif Apr 15 22:03:09 Ecco: now it won't enumerate =\ Apr 15 22:03:13 zecke: Parsing shouldn't fail ;-) Apr 15 22:03:24 RP: right, it throw an Exception on evaluating the __anonfunc() {} Apr 15 22:03:41 RP: which brings you into the generic Exception handler Apr 15 22:04:04 RP: you were not able to expand 'RDEPENDS' Apr 15 22:04:09 RP: and now I suffer ;) Apr 15 22:04:37 zecke: It should just remove any entry from the cache that throws an exception... Apr 15 22:05:00 RP: go ahead ;) Apr 15 22:05:36 Can we name this release "'N' stands for sucky monpoly" Apr 15 22:07:19 * RP has no objections :) Apr 15 22:07:19 Ecco: ok shows up. does your pxa device show up with product id a4a2 or a4a1? Apr 15 22:11:19 RP: found another bug Apr 15 22:13:39 zecke: go on :) Apr 15 22:14:29 TheMasterMind1, I don't know Apr 15 22:14:38 TheMasterMind1, in fact on pxa27x Apr 15 22:14:41 RP: you do not link/copy the 'seen overrides cache' Apr 15 22:14:46 udc is very "buggy" Apr 15 22:14:48 RP: how to proceed now Apr 15 22:14:52 in fact when it works it works great Apr 15 22:14:57 RP: I would patch some files Apr 15 22:14:59 but sometimes it just isn't enumerated Apr 15 22:15:05 RP: how to send you the patch of the patch? Apr 15 22:15:37 zecke: Will I see what you mean from a diff or not? Apr 15 22:15:46 Ecco: ah. well looking at ether.c there's a define if using pxa2xx that only uses a subset of cdc, but doesn't look like its being set Apr 15 22:16:03 hmm Apr 15 22:16:06 weird Apr 15 22:16:07 RP: I'm killing DataSmartPackage now Apr 15 22:16:07 zecke: Or should I commit this and we can take it from there? Apr 15 22:16:15 RP: probably better Apr 15 22:17:00 zecke: I have that exception handling working much better Apr 15 22:17:05 nice :) Apr 15 22:17:42 * zecke has now OpenBSD, DragonflyBSD, FreeBSD and SuSE running in his VMware server Apr 15 22:17:44 give me a few minutes to sanity check this diff, then I'll commit Apr 15 22:18:28 I will fool myself and look into one of the 'scripts' and study Apr 15 22:23:55 RP: with my fix we could see the breakage returning Apr 15 22:24:00 wait a second before comitting Apr 15 22:26:12 Zecke can you give me the link to the new oe website ? Apr 15 22:26:52 gerwinin: new website? If we have one, I don't know it Apr 15 22:26:57 okay Apr 15 22:27:41 Zecke: Shall we make one ? Apr 15 22:27:55 RP: please don't commit Apr 15 22:28:05 RP: we are back to the 'parsing' errors Apr 15 22:28:38 zecke: Due to who's changes? Apr 15 22:28:44 RP: mine Apr 15 22:29:20 RP: if you commit, all overrides from conf/bitbake.conf Apr 15 22:29:23 will be lost Apr 15 22:29:54 this leads to the fetchers not working correctly Apr 15 22:30:24 zecke: This was caused by the last change you made? Apr 15 22:30:42 RP: locally Apr 15 22:30:50 currently self._seen_overrides = copy.deepcopy(parent._seen_overrides) self._special_values = copy.deepcopy(parent._special_values) Apr 15 22:30:59 is missing in DataSmartPackage Apr 15 22:31:18 (well I have them as copy.copy in HEAD) Apr 15 22:31:37 okay these two dicts need COW as well Apr 15 22:31:47 BTW: Is cow patented? Apr 15 22:31:58 Lets hope not Apr 15 22:32:52 koen|sleep: how else do I test it :-) Apr 15 22:33:37 RP: okay, we can commit something that works now Apr 15 22:33:51 RP: but it will be way slower (on initial parse) Apr 15 22:34:07 zecke: :-( Apr 15 22:34:15 RP: (for now) Apr 15 22:34:19 zecke: We can't keep the speed gain in any way? Apr 15 22:34:21 cross your figers Apr 15 22:34:31 zecke: So I can commit? Apr 15 22:34:55 hold your fingers crossed Apr 15 22:35:20 ~900 files more to parse Apr 15 22:36:19 commit please Apr 15 22:38:12 zecke: ok, doing that now Apr 15 22:38:17 good Apr 15 22:38:31 RP: I will tmp decrease parsing performance due a deepcopy Apr 15 22:38:43 zecke: ok Apr 15 22:40:18 Ecco: strange, usbnet on linux doesn't claim the gadget as well Apr 15 22:40:50 *can't wait* Apr 15 22:41:23 zecke: nearly there Apr 15 22:43:58 zecke: Committed revision 434. Apr 15 22:44:07 congrats Apr 15 22:44:24 * zecke being jealous and puts a sleep(10) somewhere in the code Apr 15 22:44:45 zecke: Isn't that the deepcopy? :) Apr 15 22:44:55 right Apr 15 22:44:55 s/jealous/an asshole/ ;-) Apr 15 22:45:32 polyonymous: well I'm an asshole, cause I'm canadian... (southpark) actually because I'm jealous Apr 15 22:45:53 I don't know much about southpark, but I couldn't resist it! Apr 15 22:47:56 TheMasterMind1, I don't know where that come from Apr 15 22:49:28 RP: comitted my performance decreasing code Apr 15 22:49:50 RP: it should still be faster than before Apr 15 22:50:14 RP: could we add a version number to the cache this time? Apr 15 22:50:45 Ecco: i found some docs.. this kernel is set to use RNDIS so it works with windows, and that requires some experimental module enabled on my linux kernel, and explains why it doesn't work in osx Apr 15 22:50:58 haaa Apr 15 22:51:01 Cool ! Apr 15 22:51:15 i thought usbnet worked with windows though, i'm not sure why this rndis is required Apr 15 22:52:44 RP: I'm killing the dead classes now Apr 15 22:53:13 zecke: Even after the performance decreasing code, it seems quite fast :) Apr 15 22:53:38 zecke: A version number would be good and we should add that before any release Apr 15 22:53:44 RP: could be my vmware server instances slowing things down :} Apr 15 22:54:06 RP: (Number,[Dict]) Apr 15 22:54:23 RP: this way we can extend the format Apr 15 22:54:33 RP: always unpickle it to load the data... Apr 15 22:54:51 RP: I'm killing the various Dead Classes now Apr 15 22:55:32 zecke: I'll look at documenting the cache a bit. Only after that will I think about new code... Apr 15 22:55:51 awesome Apr 15 22:56:24 Zecke: did you place the survey already ? Apr 15 22:56:33 gerwinin: not yet Apr 15 22:58:04 Zecke: I will have a look tomorrow at making a plan for the website Apr 15 22:58:38 Zecke: it should be easy to maintain and it should be an add-on to the wiki Apr 15 22:58:46 meh, microsoft at it again, screwing compatibility left and right Apr 15 22:59:03 gerwinin: we need a clean seperation Apr 15 22:59:15 zecke: ok Apr 15 22:59:18 gerwinin: what goes on the website, what into the wiki Apr 15 22:59:30 Zecke: I will take care of that Apr 15 22:59:30 gerwinin: and it should be easily editable... Apr 15 22:59:36 nice Apr 15 22:59:53 Zecke: furthermore I want to make a clear plan for liasons per country Apr 15 23:00:03 gerwinin: good as well Apr 15 23:00:39 Zecke: for coding issues I am working now on the autentec driver it seems I got some usb endpoint problem Apr 15 23:00:59 gerwinin: I have never heard of that (autentec) Apr 15 23:01:19 zecke: it are those fingerprint sensors Apr 15 23:04:14 good nite gerwinin ;) Apr 15 23:09:20 RP: I have killed the dead classes Apr 15 23:09:36 zecke: ok, thanks :) Apr 15 23:09:57 hmm it is still transmitting Apr 15 23:10:27 436 Apr 15 23:10:33 zecke: That deep copy changed the initial parse from 2:45 to 4 mins :-/ Apr 15 23:10:36 okay I think we are not too far away from a next release Apr 15 23:11:27 I need to take an exam in "Design and Analysis of Algorithms" Apr 15 23:11:43 I think I showed some basic understanding here already... Apr 15 23:12:00 Lets convince a few people to test current head and see what state its in... Apr 15 23:12:04 I need to get done with uni faster... Apr 15 23:12:08 zecke: Just a bit :) Apr 15 23:12:17 zecke: This was why I did Physics... Apr 15 23:12:31 RP: I'm starting with Physics this year :} Apr 15 23:12:38 Physics for scientists Apr 15 23:13:07 I achieved what I set out to do - understood General Relativity :) Apr 15 23:13:36 RP, so, you're the third one? Apr 15 23:13:47 RP: I really get depressed about my big list of exams... Apr 15 23:13:58 * zecke hates being disabled/challenged... Apr 15 23:14:00 polyonymous: Well, I got as close as anyone ever gets :) Apr 15 23:14:07 :)) Apr 15 23:14:20 hehe Apr 15 23:15:16 Eddington should be proud of you. Apr 15 23:31:06 RP: I'm currently thinking of having a generic COW dict... Apr 15 23:31:46 zecke: I'm still confused as to how the dictonary works and what all the bits do :) Apr 15 23:32:27 RP: Don't ask ;) Apr 15 23:32:34 Well, I understand the idea of COW and the basic dictonary operations, just not what happens in all the classes :) Apr 15 23:32:59 zecke: If you tell me you can make it faster/neater, Im happy to go along with it :) Apr 15 23:33:04 RP: for each var you can have the var itself (content) and other flags Apr 15 23:33:45 RP: All I can say, I'm an asshole I can break it, without you being able to prove it (due lack of testcase) Apr 15 23:34:10 zecke: I guess I can claim the same thing with the Cache and the Rdepends handling :) Apr 15 23:34:32 zecke: One thing we can easily do now it make an unfulfilled rrecommend none fatal btw Apr 15 23:34:56 I'm going to bed now Apr 15 23:35:07 no COW for me, I will read some more in this texbook Apr 15 23:35:21 Probably wise. I'll not be long in following - maybe I'll finish documenting first Apr 15 23:35:38 RP: we need to find a solution for this kernel-abi thing Apr 15 23:36:10 zecke: Yes. Its one evil variable which I suspect should be made illegal :-/ Apr 15 23:36:18 but what to replace with? :-/ Apr 15 23:36:33 why does it end up in DEPENDS (for modutils?) Apr 15 23:36:50 oh it is a RDEPENDS Apr 15 23:36:52 It should only end up in RDEPENDS Apr 15 23:37:16 and it is legal to not have a valid RDEPENDS before your dependencies are build? Apr 15 23:37:38 So some would argue Apr 15 23:37:54 bitbake does ignore the versions in RDEPENDS anyway... Apr 15 23:38:51 or have a default variable Apr 15 23:38:59 Perhaps we add an exception handler to that piece of code - have it return None if the file isn't present Apr 15 23:39:03 UNKNOWN Apr 15 23:39:09 yes Apr 15 23:39:10 so the package gets the right version... Apr 15 23:39:31 Yes. At the end of the day, bitbake won't care about it Apr 15 23:40:31 okay cya Apr 15 23:40:53 zecke: ok, 'night! Apr 16 00:23:29 'night all Apr 16 00:48:04 Sooo.... What ever happened to using the internal 5500 system speaker for audio? :P I miss that. Apr 16 00:49:53 ? Apr 16 00:53:05 ! Apr 16 00:53:08 :P Apr 16 00:54:21 I'm assuming there was a kernel module or something in the ooooooold releases of OE and such that made the sl5500 play sound out of its internal speaker when there werent any headphones plugged in. Apr 16 00:54:36 No one elose knows what I'm talking about, so I figured you guys might :) Apr 16 00:54:54 s/elose/else/ Apr 16 00:54:56 development has shifted to kernel 2.6 and ALSA, generally Apr 16 00:55:27 yeah, the 2.6 switch is nice for everyone on cxxx, I'm sure :) ah well Apr 16 00:56:36 not especially me. it would have been nice for the 2.6 porter, but i keep bugging him to fix suspend on my c3000 and there are only 24 hours in a day, not 50 hours in a day. Apr 16 00:56:39 :) Apr 16 00:58:20 jnc, dont worry, the suspend doesnt work properly on the ipaqs either in certain cases no matter which kernel you run Apr 16 00:58:23 :P Apr 16 00:58:42 heh Apr 16 00:59:02 i wish there were a simple one or two liner patch i could do to completely disable suspend Apr 16 01:41:40 jnc: can't you select power management = n in the kernel config? Apr 16 01:52:32 my kernel-module-rfcomm has files for /etc/modutils/ Apr 16 01:52:49 i can't figure out how the ended up in the ipk. they're not in the image/ installation Apr 16 02:07:14 TheMasterMind1: that is a thought, hrm Apr 16 02:27:38 Traceback (most recent call last): Apr 16 02:27:38 File "/usr/bin/bitbake", line 29, in ? Apr 16 02:27:38 from sets import Set Apr 16 02:27:38 ImportError: No module named sets Apr 16 02:28:27 It is reported when i use bitbake nano. how to solve it? Apr 16 02:44:20 what version Apr 16 02:51:50 bitbake 1.3.2 Apr 16 02:52:52 i am trying to upgrade python to current release Apr 16 02:54:38 well, e-image-core has some problems on spitz again....locking up :-{ Apr 16 02:58:16 bitbake nano Apr 16 02:58:16 Traceback (most recent call last): Apr 16 02:58:16 File "/usr/bin/bitbake", line 29, in ? Apr 16 02:58:16 from sets import Set Apr 16 02:58:16 ImportError: No module named sets Apr 16 02:58:31 still have problem Apr 16 02:58:39 Readon: heh Apr 16 02:58:45 Readon: you need to update badly Apr 16 02:58:54 Readon: I removed that a long time ago Apr 16 02:59:47 JustinP, I update python to 2.4.3, the error still existed Apr 16 02:59:56 Readon: no, OE Apr 16 03:00:13 Readon: you need to pull and then update Apr 16 03:00:35 hmmm, at least my menu conversion script is still working in e-image ^_^ Apr 16 03:01:49 justinP i can't use scm. i can only update by release Apr 16 03:02:29 Readon: release? Apr 16 03:02:53 Readon: well, that was changed a while ago...and you *should* be able to use monotone...there's even cygwin releases Apr 16 03:02:55 JustinP, I mean release packages Apr 16 03:03:11 Readon: if you're using an OE snapshot.....get a newer one Apr 16 03:03:30 Readon: or you can go edit package.bbclass and remove some stuff Apr 16 03:05:01 Readon: I can give you a patch if you like.... Apr 16 03:07:42 I do it follow the GetStarted Apr 16 03:07:54 * JustinP sighs Apr 16 03:08:12 Readon: where did you get your OE? Apr 16 03:08:46 http://ewi546.ewi.utwente.nl/OE/OE.db.bz2, Apr 16 03:13:38 Readon: so you have the OE db and you're using monotone Apr 16 03:13:50 Readon: do a pull for the branch you're using Apr 16 03:16:42 damn it, why is e locking up Apr 16 03:17:27 the network is very slow. Apr 16 03:24:31 Readon: well, you should be updating your monotone DB Apr 16 03:24:58 Readon: and FWIW it's not the network/download that's slow with monotone, it's checking the revisions Apr 16 03:25:19 Readon: as I said I can give you a patch if you want it for a quick fix for this one problem Apr 16 03:25:25 Readon: just give me your e-mail address Apr 16 03:26:08 xydarcher@uestc.edu.cn Apr 16 03:26:40 thank you Apr 16 03:34:00 Readon: I still suggest you update your OE db Apr 16 03:34:24 Readon: monotone --db=OE.db pull monotone.vanille.de org.openembedded.dev Apr 16 03:36:02 Readon: I forwarded you the commit e-mail where I reverted the change that causes the problem you're seeing Apr 16 03:36:48 CoreDump|afk: the new altboot is great :-) It was very easy to add a "boot from HDD" option (and a kexec option ;-) Apr 16 03:37:53 CoreDump|afk: BTW, why isn't there a "boot from HDD" option built in? The hdd (hda3) is the best place to put images... Apr 16 03:42:29 JustinP, that is solved by upgrading python completed, but it is said that ERROR: Unable to parse conf/bitbake.conf (conf/bitbake.conf:398: unparsed line: 'require conf/sanity.conf') Apr 16 03:46:04 Readon: if you updated bitbake then clear your cache directory Apr 16 04:18:18 RP: sharpsl_rc_probe isn't being called for me any more..... Apr 16 05:41:02 hrm linuxfromscratch has a nice set of init scripts Apr 16 05:51:18 which website could i download current OE.db? i can't use monotone pull because of poor net Apr 16 05:52:12 believe me, i doubt network access is the limiting factor. Apr 16 05:52:26 monotone pull takes about 5-10 hours Apr 16 05:53:24 i can't afford that, our lab would cut off power every night. Apr 16 05:54:40 Readon: cut off power? that's a little odd. Just grab the latest snapshot, then pull. As long as you pull every day it shouldn't take too long to pull the newest revisions Apr 16 05:54:51 http://ewi546.ewi.utwente.nl/OE/OE.db.bz2 right? Apr 16 05:55:10 Readon: if you download new snapshots you're taking a lot of extra bandwidth from ewi that you really don't need to Apr 16 05:56:40 yes, i downloaded that, but it seems too much would be download. Apr 16 06:02:32 RP: you about? Apr 16 06:10:48 Readon: if you pull then you download *less* than if you download a whole new snapshot Apr 16 07:00:58 does it needed to cvs source ball from website when i use bitbake nano? Apr 16 07:10:53 Readon: all builds download source externally, either as a tarball from a site or from an SCM like CVS or SVN Apr 16 07:14:19 OMG, would you give some advise about xserver/kdrive? i only want to test this on our board, oe is too large for us. Apr 16 07:14:47 I have not find a proper way to get source files for that Apr 16 07:20:51 Readon: The source appears to be here: http://forums.fark.com/cgi/fark/comments.pl?IDLink=2015024 Apr 16 07:21:05 er Apr 16 07:21:07 wrong link Apr 16 07:21:26 http://freedesktop.org/wiki/Software_2fXserver Apr 16 07:21:30 Readon: just build it.....OE can build anything that it knows about. And many types of processors and "boards" are supported, although every person who comes in here seems to have one of their own.... Apr 16 07:23:33 we are using ARM7T on Hynix 7202 Reference Board, i have not find the machine config file, so i use arm9tmdi instead. Apr 16 07:25:18 johnX, does xorg contains the kdrive for framebuffer?> Apr 16 07:25:18 * JustinP knows little to nothing about such things Apr 16 07:25:35 xserver-kdrive does Apr 16 07:25:39 Readon: if you want to know where sources for things are check the bb files in OE for that package Apr 16 07:27:06 TheMasterMind1, that seems a test only desktop Apr 16 07:27:20 JustinP, ok, i find it's useful. Apr 16 07:30:23 JustinP: can it build me a sexy female companion? Apr 16 07:32:27 TheMasterMind1: haha, you are pray for a magic world. actually, i think the same as u. Apr 16 07:55:20 good morning all Apr 16 07:56:03 morninh Apr 16 07:56:06 h=g Apr 16 08:18:08 hey koen Apr 16 08:28:28 hey mithro & Cockroach- Apr 16 08:29:33 whats := do? Apr 16 08:30:44 immediate expand iirc Apr 16 08:31:27 mithro: http://bitbake.berlios.de/manual/ch02.html#id894553 Apr 16 08:33:43 Electric bill goes through the roof :P Apr 16 08:39:08 mrning all Apr 16 08:39:29 hey RP Apr 16 08:40:46 morning :) Apr 16 08:40:57 hey rm_you Apr 16 08:41:12 koen: :) Apr 16 08:41:17 RP: I know you' Apr 16 08:41:25 Is there a decent ipkg manual available? the one in the handhelds.org wiki isn't exactly one Apr 16 08:41:25 gah... stupid enter button :P Apr 16 08:41:42 RP: bvdd isn't acceptable for upstream, right? Apr 16 08:41:58 Cockroach-: that's the only on I know off Apr 16 08:42:06 hmh Apr 16 08:42:36 RP: I know you're a (the?) 2.6 person, but do you perchance remember if it was a kernel module or something that allowed the 5500 to play sound through its system speaker? back in like 3.2 of OZ? Apr 16 08:43:11 no one I've talked to even knows what I'm talking about :/ It's getting depressing Apr 16 08:43:44 isn't that speaker just a buzzer? Apr 16 08:44:06 yes... sorta. but it produces decent sound :P Apr 16 08:44:20 I have a couple of those things hooked up to a cd player and they work decently Apr 16 08:44:46 and in OZ 3.2ish it used to play audio on that buzzer when there werent headphones plugged in Apr 16 08:44:59 What's the purpose of arch.conf? does IPKG process all (regardless of filename) .conf files in /etc/ipkg dir? Apr 16 08:45:27 rm_you: I am the 2.6 kernel persion but know little about the 5500 I'm afraid Apr 16 08:45:32 and I cant seem to find an old release of OZ to go back and scope out what they were doing Apr 16 08:45:55 RP: darn :( Maybe Hent will know when he gets on :/ Apr 16 08:46:26 koen: bvdd won't be accepted upstream, correct Apr 16 08:46:39 koen: Instead we need the mutliple framebuffers patch Apr 16 08:47:38 RP: ok Apr 16 08:47:52 RP: I've got a feeling the pdaX people will go the bvdd way.... Apr 16 08:48:11 koen: Thats not actually what sashz said they'd do Apr 16 08:48:44 koen: Have you tried bitbake head yet? Apr 16 08:49:25 koen: In theory, zecke reverted some of the parsing speed up for stability so it should work so I'd be interested to hear what it does in the real world... Apr 16 08:49:39 RP: trying now Apr 16 08:49:51 koen: You can ignore the duplicate methods "errors" Apr 16 08:49:56 ah Apr 16 08:50:07 hey mickey|bbiab Apr 16 08:50:53 * RP -> back shortly Apr 16 08:51:55 RP: it's very vast Apr 16 08:52:59 http://oe.pastebin.com/662760 Apr 16 08:57:18 whays bvdd? Apr 16 09:00:27 mithro: pxa270 overlay and colorspace conversion (iirc) Apr 16 09:00:51 newer clamshells don't have a w100 anymore Apr 16 09:05:33 ahh Apr 16 09:07:41 2.64s. Nice :) Apr 16 09:08:03 RP: check the second one: 1.7s Apr 16 09:08:10 RP: but no build summarry Apr 16 09:09:05 koen: There's just one at the above url? Apr 16 09:09:06 koen, I see that you're talking about the loss of the w100. According to you, is it a bad thing ? Apr 16 09:10:11 RP: http://oe.pastebin.com/662764 Apr 16 09:10:33 Ecco: I only have a c700, so I don't really care Apr 16 09:10:39 oh, ok Apr 16 09:10:58 Bout how does the pxa fb generally compares to the w100 ? Apr 16 09:11:02 RP: don't forget to bump the BB_VERSION Apr 16 09:11:30 koen: neat. I'm not sure why the build summary disappears but that shouldn't be too difficult to fix (and is probably an existing bug rather than something I've added) Apr 16 09:12:13 I'm quite pleased with 1.7s. That's going to take some improving :) Apr 16 09:12:45 bitbake - now with sub-second parsing ;) Apr 16 09:14:07 That's really cool. How can I upgrade my bitbake copy ? Apr 16 09:14:43 Ecco: svn update Apr 16 09:14:47 oh, ok Apr 16 09:14:52 does it work now ? Apr 16 09:15:45 it seems so: NOTE: package gnumeric-1.6.3-r0: task do_build: completed Apr 16 09:20:28 lies, all lies. OE can't be that fast, its not allowed. Apr 16 09:20:30 03koen 07org.oe.dev * ra9b5462b... 10/packages/abiword/abiword-plugins_2.4.4.bb: abiword-plugins: temporarely remove gtkmathview Apr 16 09:20:38 s,OE,bitbake, Apr 16 09:23:15 TheMasterMind1: the TrollTech metric for FUDing is 'does it run on a zaurus sl5000d', so it still sucks ;) Apr 16 09:25:17 heh. running initial parse, bunch of 'method already seens' Apr 16 09:26:34 TheMasterMind1: Ignore them, they're harmless Apr 16 09:27:28 wow. intial parse about 4 minutes, subsequent about 3 seconds Apr 16 09:27:45 amazing, great job guys Apr 16 09:28:06 now just need monotone to be fast and i can actually convince people to use OE Apr 16 09:28:20 try monotone 0.26 Apr 16 09:28:28 that was about 8 times faster as 0.25 Apr 16 09:28:48 yea i've been meaning to Apr 16 09:28:54 TheMasterMind1: and for really obnoxious people we have daily tarballs and diffs at http://ewi546.ewi.utwente.nl/OE/ Apr 16 09:30:06 yea but you give people so many options and it just confuses them. Apr 16 09:30:16 heh Apr 16 09:30:42 "we want more options" [...] "we want less options" Apr 16 09:30:50 the everlasting battle against clueless users Apr 16 09:30:55 Its probably time to try and simplify getting started... Apr 16 09:30:58 yep yep yep Apr 16 09:31:31 i tried convincing the gumstix people to use oe, but it's hard to argue with 'svn co buildroot; cd buildroot; make' Apr 16 09:31:46 this c parser is coming along nicely Apr 16 09:31:55 that should reduce that 4 minutes Apr 16 09:32:01 awesome Apr 16 09:32:03 TheMasterMind1: 'wget http://www.nslu2-linux.org/Makefile ; make openslug' Apr 16 09:32:12 koen: that is a disgusting make file Apr 16 09:32:14 =P Apr 16 09:33:22 anyone looked at linux from scratch before? i really like their initscripts http://wiki.linuxfromscratch.org/lfs/browser/trunk/bootscripts/lfs?rev=7532 Apr 16 09:33:43 TheMasterMind1: I've used LFS for many years Apr 16 09:34:39 i think i'm going to use it instead of the initscripts bb in OE - its too messy and hacked together Apr 16 09:35:07 isn't OE, Linux from Scratch? :P Apr 16 09:35:07 anyway, it's almost 6am, so g'night Apr 16 09:35:11 TheMasterMind1: if you make it work, please integrate it in OE Apr 16 09:35:16 'night TheMasterMind1 Apr 16 09:35:22 yep, that's the plan Apr 16 09:35:35 i really need to update my OE soon. and push all these changes, i have a ton of them Apr 16 09:35:36 tmm1: night night Apr 16 09:36:22 hey pb_ Apr 16 09:36:29 hi mithro Apr 16 09:42:19 morning Apr 16 09:42:41 JustinP: because one of these days I'll hack up suck a script with partitions detection etc Apr 16 09:45:28 ah! There you are, CoreDump|home :) Apr 16 09:46:03 I need to ask you a question... ermm.... as soon as I remember what it was :P Apr 16 09:46:36 no hurry heh Apr 16 09:46:57 ah! do you perchance remember what it was that let the 5500 to play sound through its system speaker? back in like 3.2 of OZ? Apr 16 09:47:34 RP: http://oe.handhelds.org/cgi-bin/moin.cgi/GettingStarted Apr 16 09:47:39 in OZ 3.2ish it used to play audio on the system speaker when there werent headphones plugged in Apr 16 09:47:55 RP: it should be a bit clearer now Apr 16 09:48:10 rm_you: for this you require a kernel patch (which breaks Opies system sounds) or so I was told Apr 16 09:48:19 hrm Apr 16 09:48:54 I thought it might be a kernel hack, but I couldnt find an old OZ release to investigate (I wasnt into development back then, and OZ only goes back to 3.5.1 now :/ ) Apr 16 09:49:53 I could care less about opie's sounds.... I mean, if I'm not going to hear them anyway since I won't have headphones... :P Apr 16 09:50:39 Do you perchance know where I could find a copy of the old kernel that had that so I can scope out the patch? or possibly who might have the patch itself? Apr 16 09:50:51 morning Apr 16 09:50:53 http://www.hentges.net/misc/openzaurus/files/oz32_opie102-initrd-t9.tar.gz Apr 16 09:50:58 hey XorA Apr 16 09:51:02 awsome, thanks Apr 16 09:51:03 Ive got an old initrd, but kernel nope Apr 16 09:51:10 morning XorA Apr 16 09:51:54 RP: what do you think about updating bitbake to 1.3.4.0? Apr 16 09:52:11 hrm, found some sites that have the patch for newer kernels, might could fix it up for .18 Apr 16 09:52:45 koen: nslu2-linux is still is RDEPENDS purgatory (our autobuilds are all failing since the big RDEPENDS change a number of weeks ago). Would someone more knowledgeable than I in OE care to look at packages/meta/slugos-native.bb and identify why we are getting the error shown in http://logs.nslu2-linux.org/buildlogs/autobuild-nudi-last.txt ? Apr 16 09:53:09 Hmm - make that packages/meta/slugos-packages.bb Apr 16 09:53:35 http://logs.nslu2-linux.org/buildlogs/autobuild-nudi-last.txt Apr 16 09:53:50 ERROR: Nothing provides runtime dependency python Apr 16 09:54:00 NOTE: no buildable providers for ipkg-utils Apr 16 09:56:47 rwhitby: what does a bitbake -nv ipkg-utils say on that box? Apr 16 09:57:49 rwhitby: and does it happen with a clean build (after wiping /tmp)? Apr 16 09:58:15 koen: answers on the way ... Apr 16 09:59:01 hey zecke Apr 16 09:59:05 RP: it looks like we are building task-bootstrap-unionroot?! Apr 16 09:59:08 I know that we have recently cleared the cache on that machine Apr 16 09:59:45 rwhitby: hey **** ENDING LOGGING AT Sun Apr 16 09:59:56 2006