**** BEGIN LOGGING AT Wed Sep 25 02:59:59 2013 Sep 25 07:59:27 good morning Sep 25 08:09:10 morning all Sep 25 08:22:40 Good morning, is there a good way to point at your cross compiler from inside your bitbake recipie? Sep 25 08:27:18 BjornArnelid: ${CC} ? Sep 25 08:41:15 hello everyone :) Sep 25 08:43:39 In two weeks I'm gonna receive a board with a Freescale i.MX6 ARM Cortex-A9 CPU. I don't fully understand what you get when you build your custom image. While building can you choose the package manager you want? I could for instance use Pacman from archlinux? Sep 25 08:44:58 sadhuram: the supported package managers are rpm, ipk and deb; there is no support for pacman at this time Sep 25 08:46:26 ok, thanks. So if use deb, will I be able to use both debian and ubuntu repositories? Sep 25 08:49:03 When i try pointing at ${CC} in my bitbake recipie i get this: /bin/sh: 1: arm-poky-linux-gnueabi-gcc -march=armv5te -marm -mthumb-interwork --sysroot=/home/bjorn/yocto/poky/build/tmp/sysroots/qemuarmgcc: not found Should not cross compiler automatically be in place while building? Sep 25 09:42:16 hi anyone know of a intermidiet fix for yocto bug 3841? I can't update the core and to fix it in a .bbappend. I'm a bit lost here would appreciate any help/pointer thanks Sep 25 09:42:17 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=3841 normal, Medium, 1.4, ross.burton, RESOLVED WORKSFORME, apr-util-native do_compile fails when using ccache-native Sep 25 09:47:06 neg: if it really is that bug then cleaning should do it - bitbvake -ccleansstate apr-util-native Sep 25 09:49:40 rburton: ok than I might have a different problem, im building from everything clean and still hit the issue. Need to investigate more then thanks Sep 25 09:51:21 neg: if you can replicate on demand from clean feel free to re-open or file a new one Sep 25 09:56:31 rburton: will do, thanks for your help Sep 25 10:04:54 HI all, I have put BB_NO_NETWORK = "1" Sep 25 10:05:12 and PREMIRRORS += "${TOPDIR}/../packages" Sep 25 10:05:38 then I put my package in ${TOPDIR}/../packages, but during the build bitbake complains Sep 25 10:06:46 that it can not do fetch, as it can not do git ls-remote Sep 25 10:07:17 why does it try to git-clone before looking at PREMIRROR? Sep 25 10:12:48 any idea on this? Sep 25 10:13:23 Is not PREMIRRORS observed before starting to git-clone SRC_URI? Sep 25 10:37:32 drasko: does the recipe use git tags? Sep 25 10:38:03 drasko: bitbake always has to query the remote server if git tags are used Sep 25 10:38:21 drasko: a fix is to use full commit hashes instead of the tag Sep 25 11:02:20 I tried using core-image-minimal-initramfs.cpio.gz, but it fails for me just before 'switch_root' when trying to mount --move /dev ${ROOT_MOUNT}/dev Sep 25 11:02:45 not sure if that's the case of some bug in busybox mount or something else Sep 25 11:03:58 mount: mounting /dev on /rootfs//dev failed: Input/output error [ 15.641473] EXT3-fs error (device loop0): ext3_lookup: deleted inode referenced: 33715 Sep 25 11:27:06 If im missing recordmcount in my yocto linux kernel, is there a way to add "make scripts" to my build Sep 25 11:27:54 Or is there a base image that includes scripts already? Sep 25 11:29:58 Hi all. I have a local tarball of opkg-utils and it does nbot unpack with my open-utils.bb Sep 25 11:30:33 Any ideas where to look? bitbake -c unpack just creates directory structure, but tar.gz is not unpacked Sep 25 11:32:51 e2fsck on my ext3 image fixed the problem; core-image-minimal-initramfs.cpio.gz works like a charm, thanks Yocto developers, good job :) Sep 25 11:34:14 it would be great though to add some: command || fatal 'command failed' to have shell access in case something went wrong Sep 25 11:34:47 I might submit a patch to make next people's lives easier Sep 25 11:51:59 and don't know why but default core-image-minimal-initramfs is built without 'kernel-modules'. I add this in my bbappend, but I'm pretty sure anybody who uses that needs kernel-modules Sep 25 12:17:05 Krz_: certainly useful feedback... Sep 25 12:17:54 JaMa: around? Sep 25 12:18:09 RP: partially yes Sep 25 12:21:05 JaMa: Was https://bugzilla.yoctoproject.org/show_bug.cgi?id=5132 all down to the mkdir problem in package.bbclass? Sep 25 12:21:06 Bug 5132: normal, Medium+, 1.5.1, peter.seebach, NEW , pseudo: Different file attributes detected by buildhistory Sep 25 12:21:16 JaMa: or was that only part of the problem? Sep 25 12:22:53 only part Sep 25 12:23:32 procps still keeps changing file permissions with that mkdir/umask change Sep 25 12:24:50 JaMa: and you didn't find a reason for that as yet? Sep 25 12:25:30 JaMa: I'm planning to propose http://git.yoctoproject.org/cgit.cgi/poky-contrib/commit/?h=rpurdie/t2&id=24195503f8c1409a9dd97d56a6f21a7d8f14ce1e to fix the mkdir issue Sep 25 12:45:12 RP: I have similar patch in my queue (sorry I never sent v2 for that), so ACK and thanks Sep 25 12:47:02 RP: and I haven't found the reason for procps changes, last clue was that it's possibly from runstrip running in parallel and calling chmod from there Sep 25 12:47:52 JaMa: I wonder if its due to symlinked files being processed in parallel? Sep 25 12:48:00 RP: but with added debug output I wasn't able to confirm this theory (chmod was called in same order in 2 builds with different final permissions) but I don't know if I can trust order in logs Sep 25 12:48:27 yes something like that, because all those files are with u-a Sep 25 12:54:02 JaMa: hmm, those files are not hardlinked and not symlinks either :/ Sep 25 12:54:19 was a nice idea... Sep 25 12:55:14 ah right, it's u-a but not the symlink part Sep 25 12:55:29 but IIRC u-a renames binaries in do_package now Sep 25 12:55:57 so it can be related to u-a functionality interacting badly with pseudo or something running in paralle Sep 25 12:58:15 IRC webclient keeps crashing Sep 25 13:01:51 JaMa: but we just run the stripping in parallel, that should wait until it finishes all stripping before it continues :/ Sep 25 13:07:19 I have small tweak for this: meta/recipes-core/initrdscripts/files/init-live.sh, where may I submit the patch? Sep 25 13:18:58 JaMa: interestingly if I read back the file mode after the chmod, it all starts working Sep 25 13:19:20 seebs: this is looking very much like a pseudo bug :/ Sep 25 13:20:05 Krz-: its OE-Core so the openembedded-core mailing list Sep 25 13:30:44 basically adding 'live' to IMAGE_FSTYPES creates 2 additional images: .hddimg and core-image-minimal-initramfs. What can I do to create only second one, and stop creating .hddimg? Sep 25 13:36:45 Krz-: override IMAGE_FSTYPES for core-image-minimal-initramfs Sep 25 13:52:58 RP: I was trying to read it back in order to refresh OE cachedpath, but I was either reading it in wrong place or as you say it's bug in pseudo and not cachedpath Sep 25 13:54:08 RP: http://bpaste.net/show/135285/ this is what I was trying to do last time (more than week ago so I barely remember..) Sep 25 14:00:26 Hi all, I'd like to make local package repository to speed-up build Sep 25 14:00:58 do I have to copy all the files from downloads dir or only *.tar.gz and *.tar.bz2? Sep 25 14:01:09 There is a bunch od *.done files Sep 25 14:01:17 do these need to be copied also? Sep 25 14:18:53 JaMa: I've pasted some more info into that bug. I think there is a flaw in one of the pseudo speedup operations :/ Sep 25 14:22:23 What exactly have to be copied from downloads directory in order to have a local package repository? Sep 25 14:26:58 drasko: if you need to be selective, you don't need the .done files (assuming you are talking about using PREMIRRORS, or own-mirrors and SOURCE_MIRROR_URL) Sep 25 14:32:07 bluelightning: I am talking about SOURCE_MIRROR_URL Sep 25 14:34:10 RP: thanks a lot Sep 25 14:36:02 seebs: around? Sep 25 14:36:06 RP -- re pseudo, which bug? Sep 25 14:36:21 fray: 5132 Sep 25 14:36:23 RP, he usually doesn't get online for another 30-45 minutes Sep 25 14:39:51 I wonder if the log(s) need a more accurate counter.. Sep 25 14:40:06 it's possible for some operations for it to hit the log file at a different itme.. (shouldn't happen, but it's possible) Sep 25 14:40:46 but I'm guessing your analysis is right.. Sep 25 14:43:31 I'll get his attention when we comes in and we'll see what we can do about this.. Sep 25 14:44:03 but I'm guessing we'll have to change the FAST_OP to wait for a response from the server that it's queued up.. not that it's actually complete.. (but I'm not sure) Sep 25 14:44:50 fray: It might work if there was some kind of sync at process exit Sep 25 14:44:56 I'll put a note in the bug Sep 25 14:45:07 ya.. Sep 25 14:45:14 this is an interesting race though Sep 25 14:45:40 actions, even fast op are intended to be sequential.. Sep 25 14:48:15 fray: I'd appreciate it if you could get Peters attention on this since I think the priority needs to be upped of the bug, at least until we better understand the problem Sep 25 14:48:27 yup Sep 25 14:48:36 RP: Shouldn't "export BB_VERBOSE_LOGS=1" cause executed shell commands to be logged to log.do_*? Sep 25 14:49:47 Saur: don't you have to set that in a .conf file? Sep 25 14:50:05 Saur: I doubt its read from the environment by default, not in the default whitelist Sep 25 14:50:12 RP: Could explain why it didn't work... Sep 25 14:51:11 environment is cleared except for specific variables.. setings are expected to be primarily in a .conf file Sep 25 14:52:57 Ah, that worked a whole lot better (setting it in local.conf). :) Sep 25 15:55:31 Hi all. I did BB_GENERATE_MIRROR_TARBALLS = "1", and this generated all tarballs for git cloned packages. It also generated all mirrors in git2 dir Sep 25 15:56:16 I now removed git2 dir, and put BB_NO_NETWORK = "1", hoping that packages like this can be used : git2_git.xenomai.org.xenomai-head.git.tar.gz Sep 25 15:56:49 However, I have a bitbake complaining Function failed: Network access disabled through BB_NO_NETWORK but access requested with command git clone --bare --mirror ... Sep 25 15:57:15 like it is not capable to see that git2_git.xenomai.org.xenomai-head.git.tar.gz already exists. Sep 25 15:57:15 drasko: did you run BB_GENERATE_MIRROR_TARBALLS = "1" with a clean DL_DIR? Sep 25 15:57:26 RP: yes Sep 25 15:57:27 drasko: have you set a specific SRCREV in the xenomai recipe? or just "${AUTOREV}" ? Sep 25 15:57:55 specific SRCREV Sep 25 15:58:08 drasko: and that is a full hash and not just a tag/branch name? Sep 25 15:58:17 "head" in that tarball name looks worrying Sep 25 15:58:32 bluelightning: full hash Sep 25 15:58:49 drasko: what is the SRC_URI in the recipe? Sep 25 15:59:02 SRC_URI = "git://git.xenomai.org/xenomai-head.git;prtocol=git" Sep 25 15:59:22 These packages are actually in PREMIRRORS += "${TOPDIR}/../packages" Sep 25 15:59:24 ok, the head isn't what I thought it might be Sep 25 16:00:01 drasko: er, that is not how you use PREMIRRORS Sep 25 16:00:01 RP, no I have same behaviour for Linux Sep 25 16:00:02 ERROR: Function failed: Network access disabled through BB_NO_NETWORK but access requested with command git clone --bare --mirror git://git.rocketboards.org/linux-socfpga.git /home/drasko/yocto/build/downloads/git2/git.rocketboards.org.linux-socfpga.git (for url None) Sep 25 16:00:34 drasko: you need something like git://.* file://${TOPDIR}/../packages/ Sep 25 16:00:50 drasko: i.e. to map the git downloads to look in that location Sep 25 16:00:51 Well - I have a set of packages already in the directory delivered by the third party Sep 25 16:01:04 and I want to keep a few of my packages in the separate directory Sep 25 16:01:28 drasko: Right, but I repeat that syntax is incorrect for PREMIRRORS Sep 25 16:01:39 So I've put my packages in dir pointed by PREMIRRORS Sep 25 16:01:53 PREMIRRORS is not a directory, its a set of mappings Sep 25 16:01:53 RP: what can be used instead? Sep 25 16:02:08 drasko: see my suggestion above Sep 25 16:03:27 RP: I do not get it quite... So for every package to add file://${TOPDIR}/../packages/ ? Sep 25 16:03:46 THis does not seem very elegant to me Sep 25 16:04:42 Shoule I use PREMIRRORS at all? Is there a way to tell to bitbake that packages are in two different directories? Sep 25 16:04:46 *Should Sep 25 16:07:53 drasko: first, what do you mean by packages? Sep 25 16:08:36 drasko: you have a single directory with some git mirror tarballs in? Sep 25 16:08:51 RP: yes Sep 25 16:09:00 one dir with git tarbals Sep 25 16:09:30 drasko: ok, so calling it "packages" is *very* misleading for a start Sep 25 16:09:30 RP: I ment all recipes Sep 25 16:10:00 drasko: so look at own-mirrors.bbclass and you'll see an example of how PREMIRRORS is meant to be used Sep 25 16:10:39 drasko: what you need is a single line mapping "git://.*/.* file://${TOPDIR}/../packages/" Sep 25 16:11:00 i.e. telling it to look for git tarballs in that directory Sep 25 16:12:08 RP, so if I understood well, I should replace PREMIRRORS variable in my local.conf to be like this: Sep 25 16:12:23 PREMIRRORS += "git://.*/.* file://${TOPDIR}/../packages/" Sep 25 16:12:39 let me try this... Sep 25 16:15:09 hmmm... Still the same error Sep 25 16:16:26 drasko: check PREMIRRORS gets set to what you expect with bitbake -e Sep 25 16:17:23 Sorry, now I have some other error - missing some package... I'll have to fix this before Sep 25 16:17:45 But it looks like PREMIRRORS method you proposed is correct Sep 25 16:21:58 zeddii: tested the workaround? Sep 25 16:23:03 RP: no, still the same error Sep 25 16:23:33 how to check PREMIRRORS with bitbake -e? It seems to be a function, not variable Sep 25 16:24:34 zeddii: JaMa will yell at us soon on next world rebuild ;) Sep 25 16:24:46 ant_work. working on it! Sep 25 16:25:01 I'm being squeezed by a hardware failure and another request. but I'm at it again now. Sep 25 16:25:06 :/ Sep 25 16:26:36 zeddii: take the good part of that: progress on bug 4072 Sep 25 16:26:37 Bug https://bugzilla.yoctoproject.org/show_bug.cgi?id=4072 enhancement, High, 1.5 M4, bruce.ashfield, IN PROGRESS REVIEW , Avoid circular dependencies when building initramfs that includes kernel modules Sep 25 16:26:43 heh Sep 25 16:27:00 yah. that bug is the #1 priority now, so I'm on it. Sep 25 16:27:15 * zeddii fooled yocti, but by not mentioning the # Sep 25 16:27:20 s/but/by/ Sep 25 16:27:21 drasko: are you perhaps still inheriting own-mirrors which is setting PREMIRRORS ? Sep 25 16:27:44 bluelightning: actually, syntax is PREMIRRORS_append, not PREMIRRORS += Sep 25 16:27:51 seems to be different Sep 25 16:28:05 yes, the two are not the same Sep 25 16:29:04 bbl Sep 25 16:29:15 bluelightning: you are right - I am inheriting own-mirrors Sep 25 16:29:53 Let me try to prepend to PREMIRRORS then... Sep 25 16:30:00 drasko: I'm wondering if there is some line wrap issue in the way you're setting it. you can manually put a newline in with \n if I remember rightly Sep 25 16:30:49 hi @ all Sep 25 16:33:11 i have a question, the problem is i have a package erlang15 and a package erlang16. Now erlang-mochiweb needs erlang15 but a bblacer i have added sets the preferred provider to erlang 16 and even thogh i have set the preferred provider to erlang 15 in the recipe it wont build erlang15 it will build erlang16 Sep 25 16:34:03 zeddii: not so soon, because I'm running test-dependencies.sh on very big world, it's runnig 4 days already and ~ 7 more to go Sep 25 16:34:15 No, I am confused Sep 25 16:34:36 I just can not make bitbake to take my git2_... taball Sep 25 16:34:47 I am trying PREMIRRORS_prepend = "git://.*/.* file://${TOPDIR}/../packages/" Sep 25 16:35:17 However it is insisting in bare cloning git repo : Sep 25 16:35:20 git clone --bare --mirror git://git.rocketboards.org/linux-socfpga.git /home/drasko/yocto/build/downloads/git2/git.rocketboards.org.linux-socfpga.git (for url None) Sep 25 16:37:00 * zeddii rushes anyway. Sep 25 16:56:51 RP, that analysis looks right to me, FASTOP has a design flaw; I assumed that any given file would be being modified by a single program, for some reason. Sep 25 16:57:01 drasko: do "bitbake -e > somefile", the open somefile in an editor and see what PREMIRRORS got set to Sep 25 16:57:20 RP: It was a \n probelm, as you said Sep 25 16:57:21 Off the top of my head, possible fixes or near-fixes include: Sep 25 16:57:31 PREMIRRORS_prepend = "git://.*/.* file://${TOPDIR}/../packages/ \n \ https?$://.*/.* file://${TOPDIR}/../packages/ \n \" Sep 25 16:57:38 THis is what was needed Sep 25 16:57:54 drasko: ok, good :) Sep 25 16:57:58 1. atexit() hook which does nothing but send a ping. Since ping requires a response, this should synchronize things. Sep 25 16:58:09 Only this works, nothing elese, not even =+, not without \n Sep 25 16:58:19 very strange and confusing syntax Sep 25 16:58:32 2. Have the server actually consume all the incoming messages, then put them into a timestamp-sorted queue. This requires each message to have a timestamp with sub-second time data. Sep 25 16:58:39 is there a way to exit w/o atexit firing? Sep 25 16:59:04 Oh, several. Sep 25 16:59:10 _exit() and abort(), for instance. Sep 25 16:59:19 seebs: you don't want to know about the timestamp problems I saw in lttng ;-) Sep 25 16:59:20 seebs, for #2, what happens if the message is very delayed (heavily loaded system)? Sep 25 16:59:25 But it'd catch the really obvious cases. Sep 25 16:59:35 can you intercept abort and _exit as well? Sep 25 16:59:36 I am not sure. Sep 25 16:59:39 *thinking* Sep 25 16:59:46 RP: looking PREMIRRORS with -e is not possible Sep 25 16:59:53 result is something like this Sep 25 16:59:55 PREMIRRORS() { Sep 25 16:59:56 data.delVar('PREMIRRORS') Sep 25 16:59:56 localdata.setVar('PREMIRRORS', mirrors) Sep 25 16:59:56 localdata.setVar('PREMIRRORS', mirrors) Sep 25 17:00:04 that's all Sep 25 17:00:39 can a 'dtor' be added, will that always fire on exist of an app? or is that a library thing? Sep 25 17:00:54 exist -> exit Sep 25 17:01:25 drasko: you ran it through grep didn't you? try -C5 Sep 25 17:01:36 I do not know enough to say whether it is a safe assumption that a message which has been successfully written to a unix-domain socket will be available to the other end before the write call returns and the app can exit. Sep 25 17:02:51 * RP needs to head afk Sep 25 17:03:13 seebs: I will leave you to ponder this but it is something we're probably going to need to fix :/ Sep 25 17:03:43 Yes, yes we are. Sep 25 17:04:09 This is very similar to the failure which required the may_delete/did_delete logic. Sep 25 17:04:45 *thinks* Wondering whether I could put a serial number on database entries and make that cover for it. Sep 25 17:10:46 Few more observations: You can't guarantee a cleanup call because someone can always segfault. Sep 25 17:11:48 It might be possible to add a synchronize of some sort to wait(), but it occurs to me that the server will in principle process multiple requests from a given client before it will process some from a client with a higher client ID, and it'd be pretty common for a child process to get processed after its caller. Sep 25 17:12:47 So in theory, if client #1 is doing a wait, and then a chmod, and client #2 is doing a chmod, and then an exit, you can still (rarely) get the failure even if the wait includes a ping. Sep 25 17:13:30 It may well be that the atexit() is "good enough" because in principle if things are segfaulting or aborting, we have bigger problems. Sep 25 17:16:06 In which case I can have a patch attached to the defect for testing in Not Long, I think. Sep 25 17:16:09 ya, I'm not worried so much about segfault.. Sep 25 17:16:40 The real worry would be people calling _exit(). Sep 25 17:18:52 Not sure how often that happens, and I sort of worry about breaking things that really DID care whether they exited quickly. Sep 25 17:22:45 Anyway, updated the bug with a proposed patch, if someone who has a reproducer handy wants to see whether that helps, go right ahead. :) Sep 25 17:37:09 seebs: I'll try it Sep 25 17:41:38 newbie: wanted to add "octave" package to my image "core-image-sato", what is the package name for yocto build? Sep 25 17:46:08 git grep in meta-oe shows only this: meta-oe/recipes-devtools/swig/swig.inc: --without-octave \ Sep 25 17:46:16 :) Sep 25 17:46:43 so I presume octave is part of swig package, but you need to enable it Sep 25 17:47:04 does that make any sense? I have no idea about octave nor swig Sep 25 17:50:35 same here.. no idea.. im still figuring my way around recipes and layers.. Sep 25 17:51:08 can u suggest some methods... i am all confused here with the official documentation.. any links to simplified docs will be helpful too Sep 25 18:01:22 I know only yoctoproject.org Sep 25 18:01:26 there is documentation section Sep 25 18:01:51 IRC normally works well, but at this time most of the people are gone Sep 25 18:02:32 go to official swig website, download swig package (sources) and try to compile it first Sep 25 18:02:46 figure out if that's what you need, figure out valid ./configure flags for you Sep 25 18:02:51 and then you can start with Yocto Sep 25 18:03:10 if you don't know what/where you want using Yocto makes your position only worse Sep 25 18:33:56 does IMAGE_INSTALL -= work in Yocto? ^_^ Sep 25 18:34:16 No, but there's a _remove these days. Sep 25 18:41:59 quick question - does conditional assignment "?=" in a class get finalized _after_ the recipe using that class is parsed? Sep 25 18:42:35 for image's.. if you want to exclude a package, define: PACKAGE_EXCLUDE = "...." (note this only works in master/dora) Sep 25 18:43:01 that will allow the system to know you don't want something, and then prevent it from being installed. If the rootfs can't be generated you'll get an error that should help you figure out how to resolve it Sep 25 18:44:14 (to remove something complicated, like perl.. it will likely take many iterations) Sep 25 18:55:26 denix: i believe it's parsed while the file is being parsed. so the location of 'inherit' has some importance. Sep 25 18:55:40 i am not 100% sure, but i think that's what i have observed. Sep 25 18:56:59 denix: i think that's what this is saying to (if i read it correctly): http://docs.openembedded.org/bitbake/html/ch02.html#id398548 Sep 25 18:59:44 ndec: thanks. yes, you are correct, but my issue was that += in the recipe on top of ?= in the class actually overwrites the value from class. it's a race and _append should be used instead of += here Sep 25 19:00:30 if you have += before 'inherit' then yes, i think the ?= is ignored as the variable is set when you reach it. Sep 25 19:00:39 and yes, _append fixes that. Sep 25 19:00:49 no, even after inherit Sep 25 19:01:35 if it was before, it would be obvious Sep 25 19:01:52 denix: then, no. that doesn't make sense. Sep 25 19:02:42 well, it was also on top of the override :) so that could affect it as well Sep 25 19:03:20 so, VAR_override += "val" is not a good idea :) Sep 25 19:03:49 hehe... sorry, but too late for me ;-) Sep 25 19:04:06 i can barely get that right in the morning... Sep 25 19:05:07 yeah, sometimes with all those complex overrides and conditionals it's not obvious what precedes or overrides what :) Sep 25 19:53:01 hi all Sep 25 19:54:30 question reg. sabrelite, which dtb should I use with 3.5.7 kernel? when I use the mx6qsabrelite.dtb I get no display, when I use the mx6qsabreauto.dtb I get display but mmc fails to initialize (that's using 3.5.7 from master) Sep 25 19:55:36 I noticed that mxcfb defines are missing from mx6qsabrelite.dts, is that a bug or am I supposed to do something different here? Sep 25 20:38:28 seebs: I tried the patch and it does seem to resolve that problem Sep 25 20:40:35 Awesome. Okay, now I have *two* patches to merge into pseudo "real soon now", only I am pretty heavily loaded for a bit here. Sep 25 20:41:43 seebs: could you also put the default format string into pseudolog's help output please ;-) Sep 25 20:42:23 seebs: I had to read the source the figure out how to see more than 12 chars of the process name which is not an optimal developer workflow Sep 25 20:43:09 You people are so picky. Next you'll be telling me the thing where all usage messages are delivered by punching the user in the stomach "isn't user-friendly". Sep 25 20:52:03 you guys are doing it wrong Sep 25 20:52:11 take a page from the plan9 book Sep 25 20:52:15 an have every error be "eh?" Sep 25 20:52:19 s/an/and/ Sep 25 20:54:30 * mr_science prefers "say what?!?" Sep 25 20:54:59 or the classic "lp0 error... printer on fire?" Sep 25 20:57:07 the AIX error is good.. (trying to find the exact phrase) Sep 25 20:57:24 "Shut her down Clancy, she's pumping mud." Sep 25 20:57:45 I believe that was the result of a double fault.. even better that the first customer to 'report' it in the field was an oil driller who, umm.. got very confused Sep 25 20:57:59 heh Sep 25 20:58:40 even i'm not old enough to remember the lp0 thing, but apparently that was an actual occurance Sep 25 20:59:10 line printers throwing so much paper dust that it was literally a fire hazard Sep 25 20:59:24 so it really could be on fire... Sep 25 20:59:26 Ya, I've heard of it.. but never seen it either Sep 25 20:59:41 I have seen line printers dancing around the room when something went wrong with them though Sep 25 20:59:46 (these are the big monster ones) Sep 25 21:00:08 yup, we had one in my IV&V office for many years Sep 25 21:00:10 kept slamming the head back and forth and printing crap.. was rocking back and forth since the cart it was on wasn't locked in place Sep 25 21:00:27 sounded a lot like George Jetson's car when you fired it up Sep 25 21:00:38 heh Sep 25 21:00:48 ours had locking wheels... Sep 25 21:01:01 the printer apparently broke the locking -wheel- on the car.. Sep 25 21:01:11 it no longer locked and it started moving Sep 25 21:01:16 all four had locks Sep 25 21:01:27 must've been a later design... Sep 25 21:01:41 I can't say I remember much else about it.. I was pretty young.. Sep 25 21:01:56 my dad worked for 'Moore Business Forms', and sold paper to honeywell and such.. Sep 25 21:01:58 yeah, that was my first office after grad school Sep 25 21:02:20 I had tagged along for some reason on a late visit when this happened.. Sep 25 21:02:25 early 90s on an AF base is a lot like the 70's... Sep 25 21:02:30 I don't remember which company though.. but it was a BIG printer.. Sep 25 21:02:42 hehe.. this would have been late 80's Sep 25 21:03:01 (box in the background of the university scene in War Games is a Moore box) Sep 25 21:03:07 I had a room full of those as a kid.. Sep 25 21:03:24 they still had 2(!) operational cyber 750s when i got there Sep 25 21:03:40 used for training or? Sep 25 21:03:57 my own math dept had decommissioned theirs in 1985 or so... Sep 25 21:03:59 I programmed in uTutor courseware updates in 1994 on a Cyber.. :P Sep 25 21:04:20 fray: operational meaning launch support Sep 25 21:04:51 one for realtime processing and one for non-realtime Sep 25 21:06:18 i will take this opportunity to call you all old Sep 25 21:06:19 * Daemon404 runs Sep 25 21:06:35 get off my lawn... Sep 25 21:06:52 * fray was in highschool when he worked on the cyber.. ;) Sep 25 21:07:01 i wish i had a lawn. Sep 25 21:07:19 * fray has at least 8 acres that has to be mowed. Sep 25 21:07:24 I'll let you pretend you have a lawn Sep 25 21:07:25 i had to process field data on one all through the 80s... Sep 25 21:07:40 * Daemon404 didnt exist in th 80s Sep 25 21:07:50 undergrad/grad research assitant Sep 25 21:08:04 well i did. as a foetus. Sep 25 21:08:29 lol Sep 25 21:08:43 so you're still wet behind the ears then... Sep 25 21:10:31 indeed Sep 25 21:10:36 i made sure to start young though Sep 25 21:10:41 linux from age 11+ Sep 25 21:11:12 I started using Linux in about 1992.. I was 16.. Sep 25 21:11:31 'er.. make that 1991 was 16.. :P Sep 25 21:11:43 I started programming in 1994.. Sep 25 21:11:44 my town didnt get internet until like 1998 Sep 25 21:11:47 (linux that is) Sep 25 21:11:48 so theres that. Sep 25 21:12:01 I'd been programming since 1984-86ish.. BASIC and Assembly Sep 25 21:12:21 * fray ran a BBS from 1986 -> 1992 Sep 25 21:12:43 ahh the good 'ol days, where 1200 was 'fast', then 2400.. then 9600, then 19200, then finally 56k Sep 25 21:13:07 you can relive those to an extent Sep 25 21:13:13 all you need is a serial cable Sep 25 21:13:18 and the will to push data through t Sep 25 21:13:19 it Sep 25 21:13:30 * mr_science first punched cards in 1979 Sep 25 21:13:52 programmable calculators before that... Sep 25 21:14:08 come to canada Sep 25 21:14:14 they still need pdp11 engineers Sep 25 21:14:33 nope, just cyber/vax Sep 25 21:14:44 post-pdp11 Sep 25 21:15:08 heh Sep 25 21:15:14 i suppose vax people will be paid big $$$ for years to come Sep 25 21:15:23 but probably hate their lives. Sep 25 21:15:57 whatever i knew about the vax i've forgotten several times over Sep 25 21:16:30 haven't touched a cyber for a long time either... Sep 25 21:17:08 again im young; started on c64 BASIC Sep 25 21:18:15 heh that is young? :) Sep 25 21:18:29 im 23; thats young in here Sep 25 21:18:32 in embedded land Sep 25 21:21:48 speaking of embedded land... Sep 25 21:22:33 this must be the only place where a successful install log is filled with 99% "ECC: 4 uncorrectable bitflip(s) at offset 0x002c6000" Sep 25 22:16:16 This is a gorgeous variable expansion bug and I must share it. Sep 25 22:16:50 For various perfectly good reasons, I want to copy any files that would have been picked up by ${FILES_linux-libc-headers-dev} from one directory to another. Sep 25 22:17:19 And since at least one of those names is "/usr/include/asm*", I have to be a little clever about this, because a naive shell expansion will pick up the set of asm* directories on the host, rather than for the target. Sep 25 22:17:36 So this results in a hunk of shell code which contains "${FILES_linux-libc-headers-dev}". Sep 25 22:18:06 And if this gets built for a multilib, well. The FILES variable got renamed to FILES_lib32-linux-libc-headers-dev, so there's no longer a variable with the original name. Sep 25 22:18:17 Since there's no variable, bitbake leaves it for the shell. Sep 25 22:18:33 And the shell then interprets it as "if $FILES_linux is not set, expand to libc-headers-dev". Sep 25 22:24:10 wow Sep 25 22:24:16 I did not know that shell trick :-) Sep 25 22:25:28 ${foo-bar} => bar if foo is unset, ${foo:-bar} => bar if foo is unset or an empty string. Sep 25 22:25:45 So this only happens because the bitbake name happens not to be a valid shell variable name, but does match a shell expansion behavior. Sep 25 22:38:24 seebs: you have a chance to look deeper at that ftw() issue? Sep 25 22:39:15 Not past what I found initially: ftw() always reports targets instead of links, nftw() does unless an additional flag is being passed to it. Sep 25 22:39:21 hmm Sep 25 22:40:24 I don't see anything that should interrupt or change that. We're not, say, corrupting the flag argument in any obvious way. Sep 25 22:40:38 seebs, right but without pseudo, the sizes reported into the stat structure is a link size not the target size Sep 25 22:40:56 What call is being used to create these? Sep 25 22:41:09 Because if it's not nftw() with FTW_PHYS in the flag field, that's very surprising. Sep 25 22:41:17 And if it is, pseudo shouldn't affect its behavior, I don't think. Sep 25 22:41:33 Well, hmm. I do see one thing. Sep 25 22:42:05 Which is that the original determination of the path to pass to the real nftw() isn't reflecting any flag states. Sep 25 22:46:09 If there's an easy way to extract a trivial reproducer, I can probably try to get a look at it "soon", but that may or may not imply a particular day. Or week. Life is being exciting. Sep 25 22:53:05 seebs: yes, the producer is creating a btrfs image, which will fail, I can get you some strace output if that helps. Sep 25 22:53:34 seebs: you might be able to take the ftw() code from mkfs.c (in btrfs-progs) and write a really simple test case. Sep 25 22:55:50 Hmm. *thinks* Sep 25 22:56:07 So, they *want* the target size (which is what ftw() should give them), but they are somehow getting the link's size instead? That is odd. Sep 25 22:57:04 But I don't immediately know of any cases where pseudo will give the details of a link instead of its target for ordinary stat calls. Sep 25 23:00:20 seebs can I run a simple command under pseudo Sep 25 23:00:42 Well, in the default case: /path/to/pseudo Sep 25 23:00:50 Optionally, you can specify -P Sep 25 23:01:03 I think. Sep 26 02:39:01 How do you make bitbake download source from a Git repository when it builds a recipe? Sep 26 02:45:37 * thirtythreeforty found it, never mind **** ENDING LOGGING AT Thu Sep 26 02:59:58 2013