**** BEGIN LOGGING AT Mon May 12 02:59:57 2008 May 12 07:13:20 likewise: I'm not sure applying that autotools fix was a good idea :( May 12 07:39:16 good morning! May 12 07:48:12 looks like openssl 0.9.8g is busted, it fails at install for powerpc and at configure for i586 May 12 08:40:51 likewise: your autotools fix you nicked from OM was specifically NAK'ed by RP May 12 08:41:51 likewise: and you also damaged another commit I think May 12 08:57:48 hi, all! May 12 08:58:08 do anybody remember which of PXA3 or PXA2 CPUs contain CAN? May 12 08:58:23 or these are probably IXPs? May 12 09:00:48 slapin: I don't think its PXA May 12 09:01:37 RP, which one could you recomment except atmel, which are alredy picked up? May 12 09:01:53 slapin: I have no idea, sorry May 12 09:11:36 03koen 07org.oe.dev * r4210a3d1... 10/ (8 files in 3 dirs): gcc csl 2008q1: make it look more like other gcc recipes May 12 09:12:42 RP: revision 4b02add2b83e66d45ff3e840056487a7274d9011 merged the stuff from OM you asked me not to merge May 12 09:13:18 XorA: I know :( May 12 09:13:43 RP: it also damages the -t ${to} change, so you can revert it :-) May 12 09:13:44 XorA: The argument is going to be "OE.dev was broken" and had to be hacked to fix it May 12 09:15:52 * RP wonders what the point is anymore :( May 12 09:16:25 * XorA sends a message to email list May 12 09:16:57 We should probably remove the oe_libinstall entirely May 12 09:19:37 RP: that will break a few staging functions where the makefile is useless May 12 09:20:26 XorA: I just mean in autotools_stage_all May 12 09:22:09 RP: ah :-D May 12 09:23:08 * XorA realises he didnt eat breakfast yet and goes to find food May 12 09:24:29 Now I understand the difference between .lai and .la files, I think we can fix this properly May 12 09:31:13 hi all ! May 12 09:31:53 * steliosk things that 'Windows malicious software removal tool' is a contradiction in terms May 12 09:32:13 it would have to remove windows from the pc May 12 09:39:52 XorA: I'm thinking about something like http://www.rpsys.net/openzaurus/temp/autotools.patch May 12 09:44:31 any idea who updated openssl to 0.9.8g ? May 12 09:45:35 me May 12 09:45:49 XorA: ah :) what machine are you building for? May 12 09:46:19 jeremy_laine: its a merge from OM, so none May 12 09:47:11 XorA: it looks as though shared lib generation fails for powerpc, and for qemux86 it fails even earlier (linux-pentium config no longer exists) May 12 09:47:39 jeremy_laine: feel free to fiddle May 12 09:47:48 XorA: why thank you :) May 12 09:49:04 RP: that patch looks sane, and is less code May 12 09:49:45 jeremy_laine: sorry I gave the recipe a quick eyeball before merging, but obviously that wasnt good enough, although I would only have tested for arm anyway May 12 09:50:52 jeremy_laine : The lib generation might be because of the latest autotools.bbclass May 12 09:50:59 XorA: no worries, it was about time we picked up openssl 0.9.8 so we'll need to fix it anyway May 12 09:51:11 steliosk: I doubt it, openssl does not use autotools May 12 09:51:29 steliosk: hi by the way! May 12 09:52:01 * steliosk though it did May 12 09:52:35 steliosk: nah, it's a hand-crafted "Configure" script + makefiles May 12 09:52:49 * XorA hates them May 12 09:54:07 * broonie has one Debian package where he has to run sed on the build tree after compilation due to that stuff. May 12 09:54:14 s/build/built/ May 12 10:04:20 broonie: that sed used to be hidden in oe_libinstall May 12 10:17:09 03koen 07org.oe.dev * r352a68da... 10/ (1 packages/gcc/gcc-cross-initial_4.3.0.bb): gcc 4.3.0: add symlink hack to get the right includes to glibc May 12 10:55:28 XorA: the openssl build process is pretty twisted, it bounces back and forth between directories and it looks as though some environment variables are getting lost in the process May 12 10:57:31 workaround to get the shared libraries to build: May 12 10:57:32 - oe_runmake May 12 10:57:34 + oe_runmake all build-shared May 12 11:53:20 ~lart XorA May 12 11:53:20 * ibot overclocks XorA until XorA burns out May 12 11:53:31 ~prod ScaredyCat May 12 11:53:32 * ibot zaps ScaredyCat with a high voltage cattle prod May 12 11:53:50 you broke it didn;t you. May 12 11:54:05 nothing broke here May 12 11:54:36 that's cos you never build anyhting ;) May 12 11:55:20 http://pastebin.ca/1015674 May 12 11:56:08 that's after that monster update yesterday May 12 11:56:24 XorA: yeah, that and some more :-) May 12 11:56:27 looks like you changed SRCREV without touch conf/local.conf May 12 11:56:41 I haven't May 12 11:57:26 SRCREV_pn-libgsmd ?= "4335" May 12 11:57:31 so something is wrong :-D May 12 11:58:23 autorev May 12 11:58:56 well dont moan at me if you use random SRCREV :-) May 12 11:59:39 yeah, become commiting broken stuff works well May 12 11:59:42 because May 12 12:00:07 it wouldn;t hurt to test stuff... May 12 12:00:19 erm, Ill point to my fully tested build May 12 12:00:51 of the latest rev of libgsmd? May 12 12:01:08 of the version that is pinned May 12 12:01:41 hence my point May 12 12:02:12 so in fact your problem has nothing todo with the commit I made, you just randomly associated it? May 12 12:02:26 and hence have no real point May 12 12:03:11 no, it's always your fault. (Occasionally florian too) But always yours/ May 12 12:03:19 :-D May 12 12:03:57 lol May 12 12:04:04 where is my cat toy May 12 12:19:43 hmm, its foggy here... May 12 12:21:02 rp we have summer May 12 12:21:04 now May 12 12:21:07 in germany May 12 12:24:08 woglinde: We did over the weekend too, its just gone again... May 12 12:25:37 rp *g* did you a trip with your bike? May 12 12:26:27 woglinde: I did, yes, just into Scotland and back :) May 12 12:27:50 build a sane srcrev, clean it, then change the sane srcrev number, then building the same package will fetch the first sane srcrev.... strange May 12 13:26:22 ~curse mtn May 12 13:26:23 May you be reincarnated as a Windows XP administrator, mtn ! May 12 13:30:06 03rpurdie 07org.oe.dev * r1019be79... 10/ (1 classes/autotools.bbclass): autotools.bbclass: Revert damage from 4b02add2b83e66d45ff3e840056487a7274d9011, convert autotools_stage_all to use sed on .la files instead of calling oe_libinstall, remove broken enable-mainainer-mode flag May 12 13:33:41 ~praise RP May 12 13:33:41 All hail RP! May 12 13:37:24 ~googlefight cheese vs fake cheese May 12 13:39:52 XorA: You might want to test that for OM and bring everyone back in sync... May 12 13:40:03 RP: planning on it May 12 13:40:09 I've thrown it into Poky too May 12 13:40:18 RP: probably back end of this week Ill propagate again May 12 13:40:23 morning May 12 13:40:34 RP: or just before we go git anyway May 12 13:40:35 XorA: ok, that should give any bugs time to show :} May 12 13:40:35 ~blame florian May 12 13:40:35 * ibot blames florian (and Canada) for all the evil in the world May 12 13:40:47 RP: OM goes git that is May 12 13:41:07 XorA: ;-) May 12 13:41:46 when is poky going git? May 12 13:43:29 RP: The "damage" from commit 4b02 was proposed by zecke to repair the damage done by 85a5e1 (from OM, by xora). I hate it when the infrastructure gets broken and nobody knows why we did stuff. May 12 13:43:39 03jeremy_laine 07org.oe.dev * r07638b4b... 10/ (3 files in 3 dirs): May 12 13:43:39 openssl_0.9.8g.bb: fix generation of shared libraries (use correct CC) May 12 13:43:39 * while building the shared versions of libcrypto and libssl, the environment May 12 13:43:39 variables were lost, so we ended up using the wrong CC for linking May 12 13:44:10 like2wise: The damage was the cp -t change which was reverted May 12 13:44:30 RP: what's wrong with the cp -t approach? May 12 13:44:45 like2wise: old versions of cp don't support -t May 12 13:44:52 and we don't need it there anymore May 12 13:45:18 RP: ok, but then OM should back out as well. May 12 13:45:26 otherwise, it will creep in again someday May 12 13:45:31 like2wise: it wont May 12 13:45:35 like2wise: XorA will deal with it May 12 13:45:41 ok, tnx May 12 13:45:45 like2wise: I specifically asked RP before not merging that May 12 13:46:13 The whole change was a quick fix in OM which I asked not to be merged since I wanted to do something better in .dev May 12 13:46:13 xora: but it was your commit merging it into OE, afaics. May 12 13:46:24 like2wise: if in doubt about OM tree, feel free to give me a shout May 12 13:47:24 xora: ok, the question was about commit 85a5e185, which introduced a check for uniqueness of .lai files. May 12 13:47:55 like2wise: We agreed that should go in since it highlights big bugs in what we were doing May 12 13:47:55 like2wise: that was OK'd for merging, and it seems it did its job, it exposed problems elsewhere May 12 13:47:57 xora: I could not find into on .lai files in the libtool docs, but gettext does not pass that test so all uclibc builds were instantly broken. May 12 13:48:16 s/into/info/ May 12 13:49:03 xora: so the real question should probably be: how do we deal with packages not having unique ".lai" files in one dir? May 12 13:49:07 like2wise: does it pass now with RP's change? May 12 13:49:20 like2wise: See my last commit, the problem is dealt with May 12 13:49:32 XorA: I will build from scratch then May 12 13:49:44 I'd just prefer to have dicussed the changes rather than being forced into commiting with them ASAP May 12 13:50:10 like2wise: next time flag my commit as bad on the email list, and we shall jump to attention :-) May 12 13:51:11 We did expect some of the efl stuff to break with that change, we didn't expect gettext to have issues... May 12 13:51:21 RP, XoRa: at least for the non-package-recipe commits (so, "framework commits") it's indeed good to know what it is supposed to fix, and follow-up with some build tests. May 12 13:51:26 but the end result is good, we cound a real problem... May 12 13:51:56 RP: I never reverted koens hack to EFL May 12 13:52:02 RP: but I guess I can now May 12 13:52:17 XorA: yes, that would be good May 12 13:53:04 like2wise: well it passed all testing here, but we are not a uclibc shop May 12 13:53:10 XorA, RP: next time I will not jump to commits this fast, sure. May 12 13:54:14 XorA, RP: Any (possible) invasive change to the classes/toolchain/conf/ stuff should probably be paralleled with a heads-up email on the ml. May 12 13:54:47 like2wise: That is effectively any change to classes/ ... May 12 13:55:26 gettext builds for me May 12 13:55:29 with uclibc May 12 13:55:42 gm May 12 13:55:51 and I integrated the patches from poky May 12 13:56:07 goal is working gettext 0.17 May 12 13:56:07 morning sakoman May 12 13:56:19 Anyone had issues building minimo with latest .dev (glibc)? My build dies with "cat: ./config/build_number: No such file or directory" May 12 13:56:20 but I have to rewrite autools stuff May 12 13:56:22 woglinde: My work was useful or not? May 12 13:56:29 rp??? May 12 13:56:38 I olny saw the patches fir 0.14.1 May 12 13:56:51 no recipes for 0.17 May 12 13:56:53 sakoman: minimo is a PITA to build May 12 13:57:00 :-( May 12 13:57:05 Any advice? May 12 13:57:08 sakoman: its been a year or so since I last tried May 12 13:57:09 woglinde: I linked to http://www.rpsys.net/openzaurus/gettext/ last night May 12 13:57:44 I saw the package binary in the repo so I thought it would work ;-) May 12 13:58:09 sakoman: make sure you try the same CVS date etc of that package May 12 13:58:17 There is some old version that builds in poky May 12 13:58:31 OK, I'll take a look. Thanks! May 12 13:58:41 minimo is dead replaced with mozilla mobile, so its probably b0rk3d these days May 12 14:03:13 rp why you included mickeyls vpath patch` May 12 14:03:22 its patching Makefile.in May 12 14:03:47 woglinde: not sure, that sounds wrong May 12 14:03:57 *g* May 12 14:04:04 hm I didnt look at intl May 12 14:04:19 maybe the author is using a Makefile.in there May 12 14:04:33 okay I will go into the sun May 12 14:04:40 with my son May 12 14:04:42 till later May 12 14:04:51 woglinde: enjoy :) May 12 14:21:44 RP: still got/happy with the triumph motorcycle? May 12 14:21:44 03rpurdie * r1070 10/ (6 files in 6 dirs): utils.py: Add md5 and sha256 checksum generation functions from Ross Burton (via Poky) May 12 14:22:13 like2wise: Yes, I still have it and I'm still very happy with it :) May 12 14:22:43 RP: riding it a lot? (I'm one of those good-weather-only drivers) May 12 14:23:19 like2wise: I don't find as much time to as I'd like May 12 14:24:02 like2wise: I did all weather commuting for a few years on its predecessor even through snow on a couple of occasions and I kind of miss that... May 12 14:25:47 Having said that,I was out on the daytona in some of the most horrible weather I've been out in just a couple of months ago. The snow wasn't lying but the bike was getting blown around a lot in cross winds and the visibility was bad... May 12 14:27:00 rain, sleet, snow, cross winds and the tyres were worn out and had little tread and were race compound style ones :/ May 12 14:27:01 RP: wow, sounds like some top-notch bad-weather-flying. I hate the combo of bad weather and dense fast traffic in .nl May 12 14:27:25 like2wise: This was out in the country, so no traffic thankfully May 12 14:27:57 urban traffic is something I can't stand ;-) May 12 14:28:47 unless the roads are such you can cut through it I guess... May 12 14:29:05 RP: urban traffic is ok with me (town traffic), but not the highway traffic. too many ppl not using their eyes and mirrors. May 12 14:29:25 like2wise: What kind of bike have you got? May 12 14:29:42 RP: still my beginner's bike, a 20-year old Kawasaki GPZ500s May 12 14:30:15 Thats not so bad, it should have enough power to let you control your road position a bit May 12 14:30:58 For a couple of months I used a 125 on some major highways here and that was dangerous May 12 14:31:40 you couldn't accelerate to the top road speed easily and if you did get any speed you tended to prefer not to brake for anything... May 12 14:34:12 * mwester imagines a poor little 125 revving its little mechanical heart out, desperately trying to reach highway speeds... May 12 14:34:15 RP: lol May 12 14:34:43 RP: http://www.esrac.ele.tue.nl/~leon/gpz500_.jpg (Doesn't look too bad after 20 years of torture). May 12 14:34:56 mwester: It could manage 65mph, just never quite 70... May 12 14:37:07 like2wise: Previous to the 675 I had an 18 year old VFR400 which also still looks good :) May 12 14:37:25 The GPZ does look in good condition May 12 14:37:53 RP: funny, I always thought motorcycles would rust away much earlier than a car does. May 12 14:38:16 like2wise: http://www.rp.rpsys.net/images/bike/cadwell/cadwell2005.jpg May 12 14:38:46 like2wise: you can play spot the crash damage :/ May 12 14:39:15 RP: what's the red tape on the lights for? May 12 14:39:30 http://www.rp.rpsys.net/images/bike/cadwell/cadwell2007-1.jpg is one of my favourites :) May 12 14:39:40 RP: to prevent blinding the air traffic control tower :-) May 12 14:40:00 like2wise: Its on a track and the tape is there to hold the glass together should I crash it May 12 14:41:05 that's a curve (and photo) well taken indeed. May 12 14:42:05 the indicator lights look like to not survive any crash May 12 14:42:31 I think it owes more to the skill of the photographer than anything I was doing if I'm honest ;-) May 12 14:43:08 That bike wouldn't survive a crash very well, the indicators would be the least of my worries :/ May 12 14:43:37 Some people recommend fitting crash mushrooms and removing the mirrors/indicators for that reason May 12 14:43:49 RP: I never brought my bike across a track (apart from my manouvres for the final tests) so I can't tell :-) May 12 14:44:14 Personally, I find indicators useful to ride to the garage to refill it with fuel between sessions May 12 14:45:18 like2wise: If you get the oppertunity with a friendly group of people its worth doing for the experience. You need the right people though IMO May 12 14:45:34 As it happens I just booked up for cadwell2008 today :) May 12 14:46:35 :-) May 12 14:55:57 XorA: I have fixed the generic build error for openssl 0.9.8 (problem with shared libs generation) but x86 builds still failed May 12 14:56:09 burn x86 :-) May 12 14:56:41 XorA: the reason seems to be that we forcefully set CFLAG (no "S") in openssl.inc, totally disregarding the flags automatically set by Configure May 12 14:57:42 jeremy_laine: zecke was the guy who added that to OM repo May 12 14:57:59 XorA: more specifically "OPENSSL_BN_ASM_PART_WORDS " gets undefined, which means we end up with both an ASM and a C implementation of some functions.. May 12 14:58:23 jeremy_laine: sounds like we need different choices depending on arch, lucky we got overrides May 12 14:59:11 XorA: the problem here is that we are actually trying to be *too* smart, so I'd say that on the contrary we need fewer options.. May 12 14:59:39 jeremy_laine: well its always likely that improvements to OE or build scripts in recipes mean old hacks are no longer needed May 12 16:00:16 lol May 12 16:00:28 pgn/gif bugs in windows ce May 12 17:05:34 03gremlin 07org.oe.dev * rf6c16274... 10/ (7 files in 4 dirs): neuros osd: add machine and kernel for neuros OSD v1 May 12 19:30:54 03jeremy_laine 07org.oe.dev * r51a44a03... 10/ (8 files in 2 dirs): (log message trimmed) May 12 19:30:54 openssl.inc: fix openssl 0.9.8 builds for x86 targets May 12 19:30:54 * split do_compile() into do do_configure() and do_compile() May 12 19:30:54 * don't clobber CFLAG by exporting it: edit Configure script to set May 12 19:30:54 the machine-specific flags, and let the Configure script append May 12 19:30:54 feature-based flags May 12 19:30:56 * don't use linux-pentium and linux-ppro, they got dropped in openssl 0.9.8, May 12 20:00:05 * * OE Bug 4209 has been RESOLVED (FIXED) by cliff.brake(AT)gmail.com May 12 20:00:07 * * mono 1.2.6 fails to build May 12 20:00:09 * * http://bugs.openembedded.net/show_bug.cgi?id=4209 May 12 20:01:07 * * OE Bug 2579 has been RESOLVED (FIXED) by cliff.brake(AT)gmail.com May 12 20:01:08 * * alignment faults in uclibc + EABI May 12 20:01:10 * * http://bugs.openembedded.net/show_bug.cgi?id=2579 May 12 20:30:05 Heh May 12 20:30:24 I left my PC building world since Saturday May 12 20:30:27 Bad idea May 12 20:30:53 It was running do_configure on linux-bd-neon for 2 days now. May 12 20:31:14 Result: A 38.5 GB log.do_configure file in my work directory. May 12 20:31:21 0 bytes left on my disk. May 12 20:31:40 Thanks, Megashake! May 12 21:41:42 !oebug 4242 May 12 21:41:43 * * Bug 4242, Status: UNCONFIRMED, Created: 2008-05-09 15:10 May 12 21:41:44 * * cliff.brake(AT)gmail.com: strace does not build with uclibc_svn May 12 21:41:45 * * http://bugs.openembedded.net/show_bug.cgi?id=4242 May 12 21:41:58 * timtimred has this issue also... May 12 21:42:44 in uclibc 0.9.29.... May 12 21:44:47 bbradley, the very person.. May 12 22:01:37 timtimred: I did not push the change yet May 12 22:01:41 timtimred: pushing now ... May 12 22:02:10 cbrake_away ahh... bbradley here had a fix already :) May 12 22:02:45 timtimred: ok, I'll try to get it fixed up for good tomorrow. May 12 22:02:57 what is that nature of your fix? May 12 22:04:18 cbrake_away, timtimred, bbradley: care to ask at #uclibc? or point them to our bug #? May 12 22:07:21 ok yep we are moving our bug-stalking there briefly :) May 12 22:27:01 i wonder if things would have gone better in the creation of bitbake/oe if the metadata format allowed one to define, on the fly, a schema for the metadata being used by the file in question, instead of every variable being freeform.. then a gui would be easier to design around it May 12 22:28:22 * kergoth thinks May 12 22:37:19 kergoth: I think if you did that you'd lose one of the key features OE has May 12 22:38:12 we trying to move OE into a full fledged IDE type solution for embedded distros? May 12 22:38:26 The challenge in the gui is working out what the user wants to do. We need to realise there are several different users and those users need different UIs May 12 22:39:00 Fique: Poky has gone down the route of generating toolchains and then using an external IDE to wrap those toolchains May 12 22:39:12 The same IDE can hook onto a Poky/OE build instead though May 12 22:39:57 The fits the application developer type role but not the build system maintainer and platform integrator. Whether those roles need guis is questionable though May 12 22:41:15 RP: a given package only obeys a certain number of variables, and of those, for configuration flags, one could easily list the valid values that affect its configuration, allowing the gui to provide a dropdown May 12 22:42:11 and, it would allow you to take two metadata files from two repositories from two sources and determine if a variable behaves the same for both, or if the user would have to provide two seperate values May 12 22:44:34 Thats the GUI for the build system maintainer and platform integration guy. I'm not sure a GUI is useful to that person... May 12 22:44:58 What is perhaps more useful is being able to control multiple builds, being able to run test, bisect builds? May 12 22:45:43 Those builds may be on different servers and there could be plugable notification mechanisms for builds failures May 12 22:46:06 Thats the kind of thing that I'd fine more useful than a GUI... May 12 22:54:39 i'm just mulling over possibilities. let's say a new application developer is handed an oe environment for the distribution for their device, either before or after their integration person (asuming they even have one). They want to develop or test a new piece of code which requires that, say, it be rebuilt with glibc instead of uclibc, and rebuilt pcap with a different option enabled. imo, that should be possible to do, quickly, easily, and in May 12 22:54:39 a convenient way, for such a person, with a gui, without hacking any .bb files, and have it write out the new config which he can then hand to the build/integration person May 12 22:56:27 re May 12 23:01:32 kergoth: I agree but I think this is something that the "distro" will end up having to provide May 12 23:02:03 For example "POKYMODE=uclibc bitbake poky-image-sato" vs. "POKYMODE=glibc bitbake poky-image-sato" works now May 12 23:03:34 kergoth: thats a lot like the WinCE model May 12 23:04:04 I think the idea makes sense but as an overlay for OE, not for the core May 12 23:04:43 The core should provide any common functionality to make that as painless as possible for multiple overlay providers though May 12 23:05:07 but I should sleep... May 12 23:05:11 'night all May 12 23:05:23 night May 12 23:05:30 i think the classes or packages should be able to describe what variables affect them, and the values they obey, personally. there are lots of packages that have vars set by the global config environment and even its classes but it doesnt actually use them May 12 23:06:43 night rp May 12 23:06:43 right now someone trying to build something has no idea what vars they can set that will affect the build, other than a specific few that affect global behavior May 12 23:07:49 i want to be able to do distributed compiling somehow... :) May 12 23:08:13 anyway, not thinking about oe specifically, just mulling over general concepts about metadata used at the distribution level by the various solutions providers out there, mvista's .spec and their eclipse based ui, bitbake/oe, buildroot, debian/rules.. **** ENDING LOGGING AT Tue May 13 02:59:56 2008