| 00:01.39 | kergoth | Right, OE did that, intentionally and in a more maintainable way. It was intended to be used to construct distros. debian and the like don't provide much control over their construction, you're stuck with the packages they threw together. in addition, almost all of them require building on that distro. we needed something more flexible, which we could build from scratch, and was independent of the developer's build machine distro. |
| 00:01.52 | kergoth | of course, nowadays with qemu as far along as it is, I'd approach things differently |
| 00:02.18 | landley | kergoth: what I'm trying to do is run the OE build natively under QEMU. |
| 00:02.32 | landley | All my project does is bootstrap a native development environment on target. That's where we stop. |
| 00:02.35 | kergoth | nods |
| 00:02.46 | landley | I even broke the native build control images out into a separate repo, http://landley.net/hg/control-images |
| 00:02.58 | landley | They bootstrap other people's build environments (currently only Linux From Scratch works). |
| 00:03.19 | landley | But that's just "something you can run natively within the target without having to sit there and type stuff on the console". |
| 00:03.30 | landley | I.E. cross platform regression testing via cron job. :) |
| 00:03.52 | landley | I very much want to get OE to work because it seems a lot simpler than getting gentoo to work. |
| 00:04.07 | landley | Gentoo embedded hardwires target type into every single package file; complete list of supported target names. |
| 00:04.22 | landley | (Defeats the purpose of building natively in as host agnostic a way as you can manage.) |
| 00:04.46 | landley | I've bitched at solar about this but he's been trying to disentangle himself from gentoo development for years, and who else is there? |
| 00:05.16 | landley | (And getting gentoo to work is simpler than getting Fedora's Koji or debian's debootstrap to work...) |
| 00:05.18 | landley | Oh well... |
| 00:05.50 | kergoth | trying to build any of the major distros from scratch is a pain in the ass. the build tools require so much infrastructure, it seems, when they even exist |
| 00:05.55 | kergoth | unfortunately |
| 00:06.40 | landley | It's getting better. Red Hat used to only build on a magic build cluster at Red Hat corpoate HQ. |
| 00:06.42 | landley | Now we have koji. |
| 00:06.51 | landley | Gentoo actually has a project called catalyst that's their koji. |
| 00:06.57 | landley | I have written down somewhere the name of SuSE's. |
| 00:07.19 | landley | Debootstrap is based on installing binaries into a chroot, but it's not too far from reasonable. |
| 00:07.37 | landley | But none of its' documented and there's buckets of research and experimentation to find out what to do. |
| 00:07.49 | landley | Half the control-images project's mandate is to document all this stuff. |
| 00:08.00 | landley | But that's NOT an aboriginal linux problem. That's at 1.0 and feature complete. :) |
| 00:08.07 | landley | (Knowing where to STOP is a big deal to me.) |
| 00:08.12 | kergoth | yep, which is good, sticks to what it's good at |
| 00:08.34 | kergoth | OE assumes you're crosscompiling, I don't know that anyone has ever tried to do a full native build before |
| 00:10.02 | calculus | landley: which part of gentoo hardwires target type? |
| 00:10.14 | kergoth | one of the main reasons OE exists was to have a shared repository for distribution work for embedded systems. that's where the real value was. unfortunately desktop distros don't really collaborate much (or at all) |
| 00:10.17 | calculus | is this the KEYWORD part with arm/x86/mips/etc? |
| 00:10.25 | kergoth | the tool(s) for it really are pretty weak |
| 00:11.14 | kergoth | goes to run errands |
| 00:11.14 | landley | I'd probably lean on crosstool-ng for toolchains, except that it's such an amazing pain to set up, the C library is kind of integral to building a usable target (and about 1/3 of what breaks), we need a native toolchain on target to have a development environment, it's got 8 gazillion environment dependencies to reproduce itself so running it under the resulting environment ain't trivial (which is a basic smoketest for any system bootstrap), and things being |
| 00:11.24 | landley | But other than that... :) |
| 00:12.00 | landley | kergoth: when you say "shared repository for distribution work", do you mean targets or do you mean packages that form a root filesystem? |
| 00:12.11 | landley | Because board support and forming a root filesystem should be orthogonal. |
| 00:12.21 | landley | calculus: yup. |
| 00:12.34 | landley | And the weird incestuous stuff in what base you're choosing. I blogged about this. |
| 00:12.51 | landley | The qualcomm hexagon was claiming to be a descendant of x86 just so as not to have to re-annotate 8 gazillion files. |
| 00:13.09 | landley | Unfortunately, that would occasionally pull in x86-only packages written in assembly. |
| 00:13.29 | calculus | oh, painful for new arches |
| 00:13.46 | landley | Extremely. |
| 00:13.50 | calculus | yes, I can see that |
| 00:13.55 | landley | Also painful if you want to be architecture agnostic. |
| 00:14.21 | landley | I want to throw a toolchain in the $PATH, set my target type to "gentoo-embedded" and say "emerge package". |
| 00:14.53 | landley | It doesn't like that. My attempts to set that up manually ended with very strange portage behavior, which led me to discover every package annotated with what it had been tested on... |
| 00:22.18 | *** join/#edev itdnhr (~itdnhr@c-98-217-10-21.hsd1.ma.comcast.net) |
| 00:24.30 | landley | Sigh. Trusting a verbal assurance from my manager at polycom that they'll have a contract for me within 48 hours, vs a contract waiting to be signed from TI. |
| 00:25.03 | landley | Too many factors to balance out here to make any kind of intelligent decision... |
| 00:28.51 | kergoth | landley: both. though nowadays they're being broken up into separate repositories. they're already orthogonal file wise for the most part, but they live in the same repository. a "target" is called a "MACHINE" in oe parlance, and has a single configuration file which sets variables used by the system. such as, the capabilities of the hardware, tunings for the compiler, and in a few cases, preferred providers of some fuctionality (e.g. which k |
| 00:28.51 | kergoth | ernel recipe should be used). while you can define variables within other recipes which are target specific, its relatively rare. generally speaking, machine config defines capabilities, distro config defines policies (e.g. what package manager we're going to be using), and recipes define things to be built. package selection is done via independent image recipes. you can use any image with any combination of machine and distro |
| 00:30.35 | landley | Are the machines and the packages in different directories? |
| 00:30.38 | kergoth | yep |
| 00:30.44 | kergoth | conf/machine/, conf/distro/, recipes/ |
| 00:31.41 | landley | So how do you specify "I want to build a statically linked busybox for machine X", and how do I specify "I want to build a root filesystem with this set of packages as a chroot for the current host." |
| 00:32.01 | landley | I spent a weekend trying to answer that a month or so back. Didn't get very far. |
| 00:32.46 | kergoth | MACHINE=yourmachine in local.conf to set what you're targeting, then you'd bitbake <some recipe that provides a static busybox>. i *think* there's a busybox-static recipe. otherwise you could arrange to add -static to the compiler flags for busybox and bitbake busybox |
| 00:32.53 | kergoth | runs to the grocery store |
| 00:33.10 | landley | I'll have to try this thing this weekend. |
| 00:48.31 | landley | I just created a function called this_is_so_ugly() that lives up to its name. Let's see if it fixes the problem... |
| 01:42.31 | *** join/#edev djerome (~djerome@ip24-251-138-141.ph.ph.cox.net) |
| 02:25.43 | prpplague^2 | landley: well how did the work day end? |
| 02:25.56 | *** mode/#edev [+o prpplague] by ChanServ |
| 03:08.51 | *** join/#edev calculu5 (~calculus@gentoo/user/calculus) |
| 03:47.30 | *** join/#edev dijenerate (~dijenerat@173.225.251.175) |
| 03:50.17 | *** join/#edev rasker (~Rehan@223.130.28.6) |
| 03:55.51 | *** join/#edev calculus (~calculus@gentoo/user/calculus) |
| 04:01.12 | prpplague | dijenerate: think i can sell these in barbados? http://www.dealextreme.com/p/centipede-shaped-fridge-magnet-color-assorted-38037 |
| 04:04.07 | dijenerate | prpplague: as a form if ironic marketing, they might do well... lol |
| 04:04.18 | prpplague | dijenerate: hehe |
| 04:04.38 | prpplague | dijenerate: i can hear the hate messages already... |
| 04:05.30 | prpplague | dijenerate: you make it to the concert? |
| 04:05.31 | dijenerate | prpplague: yeah, those would come in abundance |
| 04:05.51 | dijenerate | prpplague: no, not really my kind of music |
| 04:05.58 | prpplague | dijenerate: ahh |
| 04:06.50 | prpplague | dijenerate: heard she was there for kadooment |
| 04:06.51 | dijenerate | prpplague: I'm more a killers/radiohead kind of listener |
| 04:07.01 | prpplague | dijenerate: ahh, hehe |
| 04:07.15 | dijenerate | prpplague: she was and there are photos and videos to prove it |
| 04:07.27 | *** join/#edev rasker (~Rehan@223.130.28.6) |
| 04:07.37 | prpplague | googles |
| 04:08.03 | prpplague | holy cow there are |
| 04:08.14 | dijenerate | prpplague: yup |
| 04:51.41 | *** join/#edev mbuf (~user@nat/redhat/x-xvqjlrjhigvegpkp) |
| 05:18.49 | *** join/#edev avinashhm (~avinash-h@192.94.92.14) |
| 05:27.05 | *** join/#edev clio (~andrej@85.159.109.222) |
| 05:43.22 | *** join/#edev n900evil (~Speedevil@tor/regular/SpeedEvil) |
| 05:57.18 | *** join/#edev N900ev1l (~Speedevil@94.197.127.174.threembb.co.uk) |
| 06:12.37 | *** join/#edev n900evil (~Speedevil@tor/regular/SpeedEvil) |
| 06:16.08 | *** join/#edev SpeedEvil (~user@tor/regular/SpeedEvil) |
| 06:40.41 | *** join/#edev sjs205 (~sjs205@host86-148-85-29.range86-148.btcentralplus.com) |
| 06:49.52 | *** join/#edev avinashhm (~avinash-h@192.91.66.189) |
| 06:51.37 | *** join/#edev avinashhm (~avinash-h@192.91.66.189) |
| 06:55.11 | *** join/#edev Hobbe (~chatzilla@h-254-131.a218.priv.bahnhof.se) |
| 06:56.38 | *** join/#edev avinashhm (~avinash-h@192.91.66.189) |
| 07:37.54 | *** join/#edev kirabo (~kirabo@LCaen-156-54-47-233.w80-11.abo.wanadoo.fr) |
| 08:33.05 | *** join/#edev kirabo (~kirabo@LCaen-156-54-47-233.w80-11.abo.wanadoo.fr) |
| 08:34.39 | kirabo | I've an embedded cpu, which does constant reading of a pci chip, it appears that sometimes, the chip does not respond anymore and my kernel freeze, i'm stuck at this point because i get no error message displayed. What should I look about ? |
| 08:56.56 | *** join/#edev florian_kc (~fuchs@port-217-146-132-69.static.qsc.de) |
| 08:56.56 | *** join/#edev florian_kc (~fuchs@Maemo/community/contributor/florian) |
| 09:10.42 | SpeedEvil | Add logging to the call to read the PCI chip? |
| 09:11.34 | SpeedEvil | Detect the chip not responding, and poll the various status registers to get clues? Add more instrumentation around the calls to see if you can find out which ones might cause problems? |
| 09:19.18 | madsy | SpeedEvil: Linux kernel? If so, how far do you get in the boot process? |
| 09:19.25 | madsy | Err.. kirabo |
| 09:19.57 | kirabo | madsy: It boots correctly, the device even works for some hours, then it randomly freeze |
| 09:20.05 | madsy | Which architecture? |
| 09:20.09 | kirabo | ARM |
| 09:20.37 | kirabo | And the pci chip is an ethernet bridge on which I read the RMON counters for its 14 ports every 2 seconds. |
| 09:21.04 | madsy | Enable netcat logging, for example. It's a kernel module. |
| 09:21.21 | madsy | So you get the kernel log printed over ethernet or a serial port |
| 09:21.52 | madsy | If you get a serious kernel panic/oops, then it won't get written to the local /var/log |
| 09:22.19 | kirabo | Do i need it if i'm already connected through a serial console via RS232 ? |
| 09:22.53 | madsy | No, but you might have to enable the serial log dump when configuring the kernel, and recompile |
| 09:23.03 | madsy | Not entirely sure. Google is your friend. |
| 09:23.10 | kirabo | I'm going to try this |
| 09:23.14 | madsy | There are tons of great howtos for this |
| 09:27.03 | madsy | I had to use network logging myself a while ago, when diagnosing a weird kernel oops |
| 09:27.21 | madsy | Which I in the end figure out was caused by a bug in the phonet kernel module |
| 09:28.59 | madsy | It didn't create network domains properly, and it was loaded whenever I connected my Nokia phone to the laptop. Chrome created network domains, so while my phone was connected, Chrome would crash when I tried to close it |
| 09:29.08 | madsy | Pretty obscure |
| 09:58.04 | *** join/#edev JoeLlama (~snork@unaffiliated/joellama) |
| 10:11.56 | *** join/#edev n900evil (~Speedevil@tor/regular/SpeedEvil) |
| 10:54.06 | *** join/#edev MuscleNerd (~MuscleNer@99-65-242-37.lightspeed.irvnca.sbcglobal.net) |
| 11:17.57 | *** join/#edev T0mW (~Tom@24.102.196.152.res-cmts.tv13.ptd.net) |
| 11:19.05 | *** join/#edev T0mW (~Tom@24.102.196.152.res-cmts.tv13.ptd.net) |
| 13:57.09 | *** join/#edev JoeLlama (~snork@unaffiliated/joellama) |
| 14:48.00 | landley | Blah. |
| 14:48.21 | landley | Here 'til midnight last night finding yet more ways the chip doesn't listen to the driver once it's gone into the bad state. |
| 14:48.45 | landley | Apparently if you disable bit stuffing during driver initialization, the kernel panics. |
| 14:48.53 | landley | That's nice. I wonder what bit stuffing does? |
| 14:56.01 | *** join/#edev roxfan (dunno@91.180.52.177) |
| 15:15.59 | *** join/#edev prpplague (~danders@nat/ti/x-tkbyxboceopeszjz) |
| 15:16.41 | *** join/#edev n900evil (~Speedevil@tor/regular/SpeedEvil) |
| 15:25.52 | *** join/#edev jayabharath (~jayabhara@nat/ti/x-gfbdzsvvldxtcppz) |
| 16:00.11 | *** join/#edev miltonm (~miltonm@32.97.110.51) |
| 16:16.26 | *** join/#edev dijenerate (~dijenerat@196.3.221.40) |
| 16:27.37 | *** join/#edev n900evil (~Speedevil@tor/regular/SpeedEvil) |
| 16:28.41 | *** join/#edev sndcrb (~sndcrb@dsl-hkibrasgw2-fef8de00-126.dhcp.inet.fi) |
| 16:32.43 | *** join/#edev jevin (~jevin@ec2-174-129-222-35.compute-1.amazonaws.com) |
| 16:35.45 | *** join/#edev VS_ (~VS@nat/ti/x-kwpfoxgfowjqoksq) |
| 17:00.36 | *** join/#edev sjs205 (~sjs205@host109-149-135-104.range109-149.btcentralplus.com) |
| 17:07.32 | *** join/#edev n900evil (~Speedevil@tor/regular/SpeedEvil) |
| 17:38.53 | *** join/#edev n900evil (~Speedevil@tor/regular/SpeedEvil) |
| 18:21.52 | *** join/#edev GPSFan (~kenm@64.92.145.112) |
| 18:21.53 | *** mode/#edev [+o GPSFan] by ChanServ |
| 19:14.26 | *** join/#edev methril_work (~Rafael@189.114.111.135) |
| 19:26.19 | *** join/#edev florian_kc (~fuchs@Maemo/community/contributor/florian) |
| 19:27.07 | *** join/#edev florian_kc (~fuchs@Maemo/community/contributor/florian) |
| 19:28.00 | *** join/#edev jayabharath (~jayabhara@nat/ti/x-agdtpxizrfkhatab) |
| 19:28.13 | *** join/#edev jevin (~jevin@ec2-174-129-222-35.compute-1.amazonaws.com) |
| 19:31.42 | *** join/#edev rasker (~Rehan@223.130.28.6) |
| 20:12.30 | *** join/#edev JoeLlama (~snork@unaffiliated/joellama) |
| 20:29.32 | *** join/#edev miltonm (~miltonm@32.97.110.51) |
| 20:59.54 | *** join/#edev risca (~risca@m77-219-164-73.cust.tele2.se) |
| 21:20.05 | *** join/#edev kergoth (~kergoth@ip24-251-173-232.ph.ph.cox.net) |
| 22:30.13 | *** join/#edev jevin (~jevin@ec2-174-129-222-35.compute-1.amazonaws.com) |
| 23:08.40 | *** join/#edev dijenerate (~dijenerat@66.54.118.89) |