**** BEGIN LOGGING AT Mon Dec 02 02:59:58 2013 Dec 02 08:55:16 good morning Dec 02 08:56:07 Does anyone know how to change the -c value in the mkfs.ubifs command (in run.do_roortfs)? Dec 02 08:57:43 the ubi image is not created. I found in the log the folowing error: Error: max_leb_cnt too low (979 needed) Dec 02 08:58:07 in my run-do_rootfs, the value is set to "-c 857" Dec 02 08:58:14 I have no idea where this value is defined Dec 02 09:37:52 melonipoika: i don't think there is any default value for that in OE. so it must be your distro or machine config that sets MKUBIFS_ARGS, no? Dec 02 09:38:22 ndec, ok, thanks. Will search for that Dec 02 09:44:31 morning all Dec 02 09:47:53 ndec, thanks. Yes, it was defined in the machine conf file Dec 02 10:49:45 RP: You around? Dec 02 10:50:47 Saur: yes Dec 02 10:52:54 RP: Say I have bitbake variable that a recipe depend on. Said variable is set by an anonymous python() function which in turn reads a file and sets the variable accordingly. If I change the file so the variable changes its value, this does not trigger affected tasks to re-execute. Is there any way I can make this happen? Dec 02 10:53:55 Saur: is the file always available to read? Dec 02 10:54:05 RP: Yes, it is. Dec 02 10:54:26 Saur: Change the anon python to inline python Dec 02 10:54:43 Saur: FOO = "${@some_function(d)}" Dec 02 10:54:43 RP: Basically, it is our old configuration file for our old build system that we read and turn into bitbake variables. Dec 02 10:55:49 Saur: FOO[vardepvalue] = "" would probably also work Dec 02 10:55:58 Actually, what I have now is: require ${@whatever(d)} Dec 02 10:56:29 Saur: try the second things then, inject a checksum into vardepvalue Dec 02 10:56:32 RP: Where whatever() both reads the file, generates a bb-version of the variables, and returns the name of that file... Dec 02 10:57:11 Saur: Right, bitbake is just not seeing all the other stuff its doing :/ Dec 02 10:57:22 RP: Checksum of what, the variables value? Dec 02 10:57:43 Saur: variables or the file. Something that will change when the input data changes Dec 02 10:58:27 Saur: or list all the variables explicitly in [varpdeps] or whatever that flag is called Dec 02 10:58:53 Saur: you're just doing too much out of sight of bitbake to allow it to track the variables properly Dec 02 11:00:17 Hi all, how to copy all the files from "files" directory to staging dir? Do we use globs like SRC_URI = "file://*" ? Dec 02 11:01:11 RP: Am I right in that no events at all are triggered if nothing causes a re-read of any recipes? Dec 02 11:01:52 RP: I was hoping for an event just before it would check if a recipe existed in the cache, or something like that... Dec 02 11:02:05 Saur: correct, there is no such event Dec 02 11:02:19 Saur: thinking further you could cheat and put this file in SRC_URI Dec 02 11:02:37 Saur: the checksum code will then checksum it and reparse the recipe when it changes Dec 02 11:03:01 RP: Ah, now we're getting somewhere. :) Dec 02 11:03:43 That should work :) Dec 02 11:07:38 RP: Hmm, how do I specify the right path to use in SRC_URI? I tried with file://conf/machine/axis-config/p3367.axis-config (which is what the file is currently called in the conf directory of our layer), but it didn't work. Dec 02 11:10:19 Saur: good question. You may need to set a variable in your layer.conf file and then use that here Dec 02 11:12:27 RP: Ok... Dec 02 11:12:34 I want to make an image that fits on a USB stick of a certain size, complete with bootloader and compressed btrfs. Is there an image I can use as a starting point? Dec 02 11:12:53 * repvik hasn't touched OE since 2009 or thereabouts. It's been a while. Dec 02 11:15:22 RP: Hah, it seems to be working. Thanks a lot. :) Dec 02 11:17:15 Saur: cool :) Dec 02 11:35:03 repvik: The final format is still controlled by IMAGE_FSTYPES and setting other options for the specific output types. The image contents is unlreated to the output format Dec 02 12:17:35 If I want to add a package to my build I add it to my local.conf using IMAGE_INSTALL_append but where do I find the name of the package, in particular NetworkManager, I know where the bb file for it is but not where to find the package name? Dec 02 12:22:51 D4v33: you can always go to poky and do: "grep network ./ -r" Dec 02 12:23:15 D4v33: if that gives you too much results then make your search more detailed Dec 02 12:24:36 D4v33: one way is to look under "packages-split" in the workdir for the recipe Dec 02 12:24:58 where the workdir can be found using: bitbake -e recipename | grep ^WORKDIR= Dec 02 12:25:09 (assuming you haven't enabled rm_work, that is) Dec 02 12:25:25 in the easy case, reading the recipe will tell you what the packages are. if PACKAGES isn't set, its the name of the recipe. Dec 02 12:25:34 yep, that too Dec 02 12:32:20 I've not knowingly enabled rm_work. where in the recipe will it state the name I'm assume we are are talking about networkmanager_0.9.2.0.bb? I was hoping there was some logical connection from the package name to to the recipe? Dec 02 12:32:49 D4v33: the general logic connection is that the primary package is the name of the recipe Dec 02 12:32:58 if that isn't the case, the recipe will say so by setting PACKAGES Dec 02 12:33:32 libraries are a bit special in that they get renamed to their library name Dec 02 13:48:47 RP: Yeah, I have an idea about how it works (unless OE has changed fundamentally since 2009). Do you know if there is an image that uses grub? It's probably a lot easier to adapt something to my needs :) Dec 02 13:53:43 guys to enable remote debuging with eclipse do I neet tools-debug or eclipse-debug? Dec 02 13:53:58 on adt-manual it says to use tools- Dec 02 13:54:08 but I do not have the tcf Dec 02 13:54:19 -agent in rootfs Dec 02 14:00:05 * LetoThe2nd would say both :) Dec 02 14:01:01 then make sure the image actually gets rebuilt, so tcf-agent is really there. Dec 02 14:05:55 LetoThe2nd: I don't see the agant file in the rootfs actually. I'll do a clean Dec 02 14:10:09 repvik: I know the "live" images install to a device and include grub Dec 02 14:10:31 repvik: its not quite what you want but similar in some ways Dec 02 14:13:41 hi, where can I find the linux kernel log in my layer? Dec 02 14:13:48 ./tmp/log does not seem to contain it. Dec 02 14:18:51 lpapp: what is the linux kernel log? Dec 02 14:20:24 ndec: anything about bitbake linux-foo Dec 02 14:21:01 so, it's like any other recipe, in tmp/work////temp Dec 02 14:21:14 for the kernel it will be in , not Dec 02 14:21:31 the problem is that there are so many folders in tmp/work Dec 02 14:21:38 that always confuses me. Dec 02 14:21:56 I really do not understand why when I am only trying to build with one scenario..... Dec 02 14:21:59 it seems a bit bloated to me. Dec 02 14:22:29 then you can use bitbake -e linux-foo |grep ^WORKDIR Dec 02 14:22:30 can I eventually have one folder there? Dec 02 14:22:45 I do not understand why I would have more than one folder there..... Dec 02 14:22:53 where? Dec 02 14:23:03 currently, I have _5_, including x86_64-linux (!). Dec 02 14:23:10 you have 1 'WORKDIR' for each recipe that was built. Dec 02 14:23:21 I am talking about workdir. Dec 02 14:23:23 not* Dec 02 14:23:30 I am talking about 5 different folders Dec 02 14:23:34 in tmp/work Dec 02 14:23:39 when I only build for _one_ scenario. Dec 02 14:24:03 the folders are organized 'logically', by . you never have 1 scenario only.. since you at least have the arch, the machine and the host. Dec 02 14:24:25 sounds really bloated. Dec 02 14:24:33 evenutally, I always have only 1 scenario. Dec 02 14:24:44 eventually* Dec 02 14:24:56 is there an option to organize it nicer? Dec 02 14:25:17 it is really hard find stuff if it is this over-engineered even for the common cases. Dec 02 14:25:39 sure.. WORKDIR is set in bitbake.conf as Dec 02 14:25:39 WORKDIR = "${TMPDIR}/work/${MULTIMACH_TARGET_SYS}/${PN}/${EXTENDPE}${PV}-${PR}" Dec 02 14:26:10 right, so my image should probably override WORKDIR Dec 02 14:26:16 not to depend on millions of factors. Dec 02 14:26:19 Hi all, how to copy all the files from "files" directory to staging dir? Do we use globs like SRC_URI = "file://*" ? Dec 02 14:26:27 at your *own* risks, though.. Dec 02 14:26:53 that does not quite sound good... :( Dec 02 14:26:59 your image will still involve building 'stuff' for the host, for the target arch, and for the target machine, you won't remove that. Dec 02 14:27:37 lpapp: honestly, i think it's less troubles to learn about OE and its policies that trying to change WORKDIR... but maybe i am wrong.. Dec 02 14:28:06 drasko: you meant copy from 'files' to WORKDIR, i guess. Dec 02 14:28:17 ndec, exactly Dec 02 14:28:21 ;-) Dec 02 14:28:35 ndec: I do not wanna learn this messy stuff Dec 02 14:28:37 lpapp: no, you should not be changing WORKDIR Dec 02 14:28:40 I wanna change it to be simpler for me. Dec 02 14:28:42 i don't think you can have wildcards here. Dec 02 14:28:46 it is a terrible mess currently. Dec 02 14:28:52 I never know where to look for stuff. Dec 02 14:28:52 I have a bunch of files in the local treem, I do not want to copy them one by one. Dec 02 14:28:55 lpapp: it's not a mess, it's highly organised in fact Dec 02 14:28:59 How can this be avoided? Dec 02 14:29:15 and I have not seen any build system yet organizing stuff based on 5-6 factors at least. Dec 02 14:29:37 drasko: file:// does not imply that bitbake will take the files in './files' in fact, as it is searching in many other folders too, so a glob pattern cannot be used. Dec 02 14:29:57 ndec, what should I use then? Dec 02 14:30:00 http://pastebin.kde.org/pedqujkr8 -> in addition, there does not seem to be any compilation output. Dec 02 14:30:09 bluelightning: then we will need to disagree, I believe. Dec 02 14:30:16 drasko: i don't know if there is a better way than listing all the files... Dec 02 14:30:30 ndec, this is crazy... Dec 02 14:30:46 there is really a tons of files in my driver Dec 02 14:30:57 drasko: i am not saying you have to do that.. just saying i don't know how to do it -) Dec 02 14:31:16 drasko: wildcards are sort of supported... what kind of files are these though? Dec 02 14:31:22 you might be able to use a python snippet to list files from dir... Dec 02 14:31:52 lpapp: well, feel free to experiment if you like, but expect breakage Dec 02 14:32:07 bluelightning, out-od-tree kernel driver for wireless device: https://github.com/lwfinger/rtl8192du Dec 02 14:32:09 bluelightning: how do wildcard work then? since bitbake looks into multiple folders Dec 02 14:32:16 * lpapp will try to override WORKDIR Dec 02 14:32:28 bluelightning: it is already broken. Dec 02 14:32:35 lpapp: it is not Dec 02 14:32:36 as I wrote, there is no any compilation output, etc. Dec 02 14:32:40 sure, it is. Dec 02 14:32:48 lpapp: changing WORKDIR will not fix that Dec 02 14:33:40 bluelightning: it is not about WORKDIR; it is about that I cannot already achieve my target, so perhaps it will be simpler with less complication. Dec 02 14:33:47 lpapp: also, those are setscene logs; there is no compilation log because it hasn't been compiled, it was restored from shared state Dec 02 14:34:09 bluelightning: to me, depending on 5-6 factors with a simple build folder is over-engineering, but I guess we will have to agree to disagree there. Dec 02 14:35:17 bluelightning: it /was/ compiled. Dec 02 14:35:27 I saw with my eyes. :) Dec 02 14:35:44 drasko: I'm not quite following; are you trying to avoid pointing at the repository by copying the source next to the recipe? Dec 02 14:36:21 eventually, I even ran bitbake -c cleansstate linux-foo Dec 02 14:37:01 lpapp: then this cannot be the directory corresponding to the version of the recipe that was built, because this one definitely wasn't compiled Dec 02 14:37:18 if it were, there would be log.do_compile Dec 02 14:37:19 bluelightning: maybe a bug in bitbake, I do not know. Dec 02 14:37:25 I doubt it Dec 02 14:42:42 bluelightning: then I need your blessing, maybe. :D Dec 02 14:42:49 and some benediction Dec 02 14:43:01 clearly, I am running, what I am supposed to do, and the expected behavior does not meet the reality. Dec 02 14:43:37 but let me know if bitbake -c cleansstate along with clean is not enough before rebuilding...... Dec 02 14:44:14 lpapp: are you using SSTATE_MIRRORS ? Dec 02 14:44:41 bluelightning: I am not sure what you mean by that question. Dec 02 14:44:58 lpapp: it's a simple question - have you set this variable in your configuration? Dec 02 14:45:07 _what_ configuration Dec 02 14:45:13 it might be simple for you, but not simple for me. Dec 02 14:45:17 local.conf, distro config... whatever Dec 02 14:45:20 distro, machine, local, etc? Dec 02 14:45:56 if you're not sure, just use bitbake -e | grep ^SSTATE_MIRRORS= to check Dec 02 14:46:26 empty Dec 02 14:46:31 ok Dec 02 14:46:34 just checking Dec 02 14:47:14 bluelightning, no, I have relatively different source that resembles to this one Dec 02 14:47:22 then as long as you're using bitbake -c cleansstate on the right recipe, then it could not be rebuilding that recipe from shared state Dec 02 14:47:36 drasko: but you're attempting to put the source next to the recipe? Dec 02 14:47:42 bluelightning, yes Dec 02 14:47:54 source is located in files directory next to the recipe Dec 02 14:48:01 bluelightning: so it is a bug if it does then? Dec 02 14:48:03 drasko: I'd suggest not doing that; put the source in a git repo or tarball somewhere else, and have the recipe file fetch that Dec 02 14:48:16 bluelightning, why? Dec 02 14:48:36 drasko: because it avoids problems like the one you're having... Dec 02 14:48:57 bluelightning: to be honest, I could manually delete the build folder as usual to solve mysterious issues Dec 02 14:48:58 it is very impractical for me, Dec 02 14:49:03 but I do not have 2 hours this time if possible...... Dec 02 14:49:17 to traball every time I cahange something Dec 02 14:49:18 lpapp: when you do -c cleansstate it *deletes* the sstate packages corresponding to the recipe; there is nothing to restore from in that case Dec 02 14:49:25 this code is in the course of development Dec 02 14:49:31 bluelightning: yet, it seems to do as you claimed. Dec 02 14:49:36 drasko: then you should use a git repository Dec 02 14:49:39 what more to do to get it right? Dec 02 14:49:51 and I do not have git repo st my disposal Dec 02 14:50:10 lpapp: something else is going on that has not been brought out in what you have said, because all that you have said does not make any sense Dec 02 14:50:23 at least one that will be accesible to the people who will eventualy use this recipe Dec 02 14:50:33 so I prefer having all files there, localy Dec 02 14:51:02 so, wuestion is - is globing possible, or I have to give an explicit list of all files? Dec 02 14:51:12 drasko: you can do it, but the problems you are having are a direct result of this kind of usage not really being supported Dec 02 14:52:09 bluelightning: you re welcome to tell me what is missing..... Dec 02 14:52:17 I know it does not make sense what I get. Dec 02 14:53:20 lpapp: have you checked you are looking in the right place, by querying the workdir using: bitbake -e recipename | grep ^WORKDIR= Dec 02 14:55:34 drasko: if you insist on doing it this way you probably need to put the files in a named directory somewhere in FILESPATH and then point to that directory; note that files itself as the directory wouldn't be appropriate since that itself is in FILESPATH Dec 02 14:55:42 drasko: files/something should work though Dec 02 14:57:11 drasko: I'm not sure if you'd put file://something or file://something/* in SRC_URI though, you'll have to test Dec 02 15:04:13 bluelightning, I am looking for a methoud to create git after all... Will be better, as you suggested. Dec 02 15:05:29 drasko: great... if nothing else, it does provide some versioning advantages in terms of being able to tell which version of the source corresponds with what's installed in the image Dec 02 15:06:01 (assuming you do PV = "x.y+git${SRCPV" in the recipe that is) Dec 02 15:06:07 er missing closing } Dec 02 16:17:22 Anyone else seeing this ?, http://pastebin.com/qe578V3r Dec 02 16:29:25 poky/master b.t.w. Dec 02 16:50:19 dany: no, but there were changes in chrpath recently Dec 02 17:20:41 bluelightning: re, still around? Dec 02 17:21:16 bluelightning: I looked into the right folder, now I have this, log.do_cleansstate log.do_cleansstate.9009 log.task_order run.do_cleansstate.9009 Dec 02 17:29:39 bluelightning: http://pastebin.kde.org/po8ch2jgp => this is the list now. Dec 02 17:30:15 bluelightning: seems it got it now. Dec 02 17:31:20 lpapp: right, that looks more like it should after an actual build as opposed to a restore from sstate Dec 02 17:31:39 bluelightning: I did not run -c clean this time Dec 02 17:31:41 just cleansstate Dec 02 17:31:46 I wonder if that makes a difference. Dec 02 17:32:24 bluelightning: is there a way to get more verbose log than in log.do_compile? Dec 02 17:32:30 it does not seem to build my driver Dec 02 17:33:39 bluelightning: and it is in Kconfig Dec 02 17:34:01 bluelightning: and I enabled it with bitbake virtual/kernel -c menuconfig Dec 02 17:39:35 bluelightning: is that the file I should check? linux-foo/3.2.1-r3/linux-3.2.1/.config Dec 02 17:40:53 or which file is bitbake virtual/kernel -c menuconfig supposed to modify? Dec 02 17:47:55 lpapp: is your driver enabled in that file? Dec 02 17:51:17 bluelightning: no Dec 02 17:51:22 bluelightning: I checked that. Dec 02 17:51:30 bluelightning: I think I need to extend the defconfig in files. Dec 02 17:51:56 lpapp: did you run menuconfig just before the build or was there a cleasstate in between times? Dec 02 17:53:14 bluelightning: not sure what you mean, but the steps are: 1) Clean sstate 2) menuconfig 3) build Dec 02 17:54:21 lpapp: well, that is supposed to work, but it's not something I personally check on a regular basis... which branch are you on at the moment? Dec 02 17:55:52 bluelightning: no branch Dec 02 17:56:46 lpapp: well, clearly it's based on an existing branch at some point, so what is it based on? Dec 02 17:56:52 brb Dec 02 17:57:00 is there a mechanism in Yocto to pick up the defconfig from the files/ folder (patch basically) and modify that in the background based on what I do in "make menuconfig"? Dec 02 17:57:27 or I need to generate the defconfig on my desktop and pass it to Yocto? Dec 02 17:57:32 the former would be lotta nicer. Dec 02 17:59:00 I still do not know what would guarantee to pick up my kernel with "bitbake virtual/kernel -c menuconfig" as opposed to the one in /meta? Dec 02 18:00:45 lpapp: as far as I know it's supposed to work as you described Dec 02 18:01:11 bluelightning: can you answer my last question? Dec 02 18:01:46 I thought I did Dec 02 18:02:12 bluelightning: perhaps you did not see my last question Dec 02 18:02:15 17:54 -!- _alex_kag_ [~alex_kag@178.126.129.180] has quit [Quit: Konversation terminated!] Dec 02 18:02:18 17:54 < lpapp> I still do not know what would guarantee to pick up my kernel with "bitbake virtual/kernel -c menuconfig" as opposed to the one in /meta? Dec 02 18:02:21 oopsie... Dec 02 18:02:47 bluelightning: what did you reply to by "as far as I know it's supposed to work as you described" Dec 02 18:03:54 as far as I know it's supposed to base what is shown in menuconfig on the supplied configuration (defconfig, config fragments, whatever), and then whatever is saved in menuconfig assuming you run it should be used in the build Dec 02 18:04:08 if that is not how it works then I think that would be a bug Dec 02 18:04:21 bluelightning: but my question is more than that. Dec 02 18:04:33 bluelightning: I would like to modify the defconfig, not just the temporary build. Dec 02 18:04:48 or do I really need to do this separately from Yocto as Yocto does not provide a way for it? Dec 02 18:05:33 are you building a custom kernel recipe you've provided? Dec 02 18:06:16 no, I have a fork Dec 02 18:06:26 it is very risky to update the kernel with Yocto every half a year Dec 02 18:06:33 lpapp: so you're building linux-yocto then? Dec 02 18:06:38 no Dec 02 18:06:44 we have the version from denzil times IIRC Dec 02 18:06:45 3.2.1 Dec 02 18:06:52 so we copied it when we moved to dylan Dec 02 18:06:56 but not linux-yocto? Dec 02 18:07:13 why do you keep repeating linux-yocto? Dec 02 18:07:37 we only have linux-yocto as a functional kernel recipe in OE-Core/Poky, no other Dec 02 18:07:47 so if you're not building that then the answer to the first question should have been "yes, I am building a custom kernel recipe" Dec 02 18:08:01 let us see Dec 02 18:08:16 surely, there is meta/recipes-kernel/linux Dec 02 18:08:28 (that is not meta-yocto) Dec 02 18:09:15 no, it isn't... meta-yocto is a layer, linux-yocto is a recipe; they are not the same thing Dec 02 18:09:49 exactly what I am saying. Dec 02 18:09:53 we do not care about meta-yocto Dec 02 18:09:58 it is an OE based distribution, not poky. Dec 02 18:11:38 OE-Core has linux-yocto; meta-yocto/poky have nothing to do with this Dec 02 18:12:25 sorry, I do not know what you mean then. Dec 02 18:12:58 right, it must have been linux-yocto, I think. Dec 02 18:13:04 although it is quite confusing if you ask me. Dec 02 18:13:12 people will associate to meta-yocto Dec 02 18:14:40 so out of the box, linux-yocto doesn't use a defconfig, it uses config fragments Dec 02 18:14:55 I'd suggest having a read of our kernel & BSP manuals for further information Dec 02 18:15:10 you can, if you want to, supply a plain defconfig in SRC_URI Dec 02 18:17:40 as I said, we have been using defconfig for ages Dec 02 18:17:43 and it has been working. Dec 02 18:17:52 but I would like Yocto to give me the convenience to modify that on the fly Dec 02 18:17:59 without going to desktop and regenerate it there, meh. Dec 02 18:19:59 you can just copy the .config on top of the defconfig after modifying it with -c menuconfig Dec 02 18:20:14 I believe that workflow is described in the kernel manual Dec 02 18:20:23 but as I wrote, even the .config is wrong Dec 02 18:20:26 it does not contain my selection. Dec 02 18:20:36 also, it should be automatical for a "Save?" question Dec 02 18:20:52 we don't modify our inputs, I've mentioned that before Dec 02 18:21:01 not within the build system itself Dec 02 18:21:46 as I already wrote .config is not modified either which is temporary. Dec 02 18:22:12 right, and as I already wrote as well, if that doesn't work it sounds like a bug to me Dec 02 18:23:00 :( Dec 02 18:25:16 lpapp: so is your current branch based on denzil? or not anymore? Dec 02 18:25:45 bluelightning: it is dylan Dec 02 18:33:35 bluelightning_: I still do not know how one can select between two kernel versions. Dec 02 18:33:47 meta/ has one, so does our distro layer. Dec 02 18:33:57 how does virtual/kernel know which one to pick up? Dec 02 18:34:24 PREFERRED_PROVIDER_virtual/kernel Dec 02 18:34:34 bluelightning_: where? Dec 02 18:35:04 lpapp: what does bitbake -e | less tell you about where it's set? Dec 02 18:38:52 bluelightning: ah, it is both in the machine and distro conf Dec 02 18:57:53 bluelightning: what greatly surprises me is that I am not getting the same search result if I do the search on the desktop, and in Yocto. Dec 02 18:58:00 and the patch should be applied in Yocto. Dec 02 18:58:18 bluelightning: i.e. the Kconfig should be the same in this sense. Dec 02 18:59:03 bluelightning: maybe the patching was not successful as the desktop has the latest staging. Anyhow, shouldn't Yocto raise an error then? Dec 02 19:00:54 based on the content, the patch should have applied fine... then I have no clue why entering "/" in makemenuconfig and typing the search word results different output screen on the desktop, and in Yocto..... Dec 02 19:03:10 * lpapp will need to read https://www.kernel.org/doc/Documentation/kbuild/kconfig-language.txt Dec 02 22:20:39 kergoth: I seem to remember you doing something with json in bitbake, what was that again? Dec 02 22:24:12 RP: https://gist.github.com/kergoth/2788410 Dec 02 22:25:34 was just a prototype/experiment, playing around Dec 02 22:25:39 though it did work surprising well Dec 02 22:28:15 kergoth: right, I remember you saying that at the time. I'm wondering whether something more like our shell/python function markers would make sense for it Dec 02 22:28:56 so you could do something like json BAR {\n\n} Dec 02 22:29:12 would then give us an interesting way to get multline more complex data Dec 02 22:29:32 our current MIRRORS syntax sucks for example Dec 02 22:32:45 kergoth: I guess the piece I'm not remembering is whether ConfHandler handles multiline expressions gracefully or not. I'm suspecting the latter :/ **** ENDING LOGGING AT Tue Dec 03 02:59:58 2013