**** BEGIN LOGGING AT Thu Mar 13 02:59:57 2008 Mar 13 11:14:20 I have some questions regarding compilation of optware packages.. would this be the right forum for such inqueries? Mar 13 11:25:22 re Mar 13 11:51:23 Anyone experienced in building optware packages around? Mar 13 11:51:50 mandersson: I can't say that I am experienced in that but I did build some... Mar 13 11:52:07 what to you need? Mar 13 11:53:10 I'm trying to build mediatomb, since the current package is crippled imho, but it is linked against libstdc++ and not uClibc++... Mar 13 11:53:24 hehe, small world Mar 13 11:53:29 I'm using the super-makefile and all... Mar 13 11:53:41 what target are you building for? Mar 13 11:54:05 brcm-24 Mar 13 11:54:15 I have a wl500gP Mar 13 11:54:37 Running openwrt Mar 13 11:54:49 All I need is mediatomb with tag-support... Mar 13 11:54:54 well here you go - the openwrt uclibc version does not support exceptions Mar 13 11:55:00 mediatomb can not run without c++ exceptions Mar 13 11:55:28 Well, it's up and running with the package from the repository... Mar 13 11:55:31 and if you are not running optware but running openwrt then I am not sure if an optware package will actually work if you add it to openwrt Mar 13 11:55:42 And it is not linked against stdlibc++... Mar 13 11:55:59 where did you install mt from? Mar 13 11:56:07 from what feed? Mar 13 11:56:59 optware/openwrt-brcm24/cross/unstable Mar 13 11:57:44 aha.. I did not know that there are optware packages for openwrt, that's cool Mar 13 11:57:46 An earlier release was built with tag-support but not with PS3-support and I need both :) Mar 13 11:58:29 so the latest package, 0.11.0-4 I think - is without taglib? Mar 13 11:59:04 yes Mar 13 11:59:15 Neither taglib nor id3lib Mar 13 11:59:31 sec, phone Mar 13 11:59:35 k Mar 13 12:00:04 Apparently someone is able to build a functional binary, I want to know how :) Mar 13 12:04:19 re Mar 13 12:05:53 well you can try the following Mar 13 12:06:00 check out optware Mar 13 12:06:11 make wl500g-target Mar 13 12:06:16 cd wl500g Mar 13 12:06:21 make directories Mar 13 12:06:23 make toolchain Mar 13 12:06:24 make mediatomb Mar 13 12:06:55 hmm, make toolchain does not do what I would expect :) Mar 13 12:08:29 The configure-script fails for the wl500g-target.. Mar 13 12:08:48 I think that it is for ppl with olegs firmware or something.. Mar 13 12:09:37 well, I happen to be one of the meidatomb devs and I plan to build static binaries for various devices Mar 13 12:10:04 so that those could be used on any linux based firmware if necessary Mar 13 12:10:15 Ah, I see. Mar 13 12:10:16 Oh Mar 13 12:10:22 That would do. Mar 13 12:10:40 but Optware is pretty cool, a lot of devices are supported Mar 13 12:11:11 Yeah, and apparently someone is sitting on a toolchain that manages to build mediatomb correctly.. Mar 13 12:11:56 Oh well. Mar 13 12:12:08 the version that worked for you - was it also from the optware feeds? Mar 13 12:15:05 Yes. The first one was 0.11.0-1 I believe but it lacked PS3-support. Mar 13 12:15:18 Or if it was -2, can't remember. Mar 13 12:15:21 I doubt that, PS3 support must be enabled in config.xml Mar 13 12:15:29 and it does not have external dependencies Mar 13 12:15:37 But there is a configure-flag right? Mar 13 12:15:39 so PS3 should work regardless of those issues Mar 13 12:15:45 yes but it is "on" by default Mar 13 12:15:56 *doh* Mar 13 12:17:30 Now, if I only had a time-machine to fetch the old package.. Mar 13 12:17:57 well, eno might be around later on, he will probably have an idea on what went wrong Mar 13 12:18:23 Great. Mar 13 12:19:12 It would be nice to be able to built it myself as well though. Mar 13 12:19:18 Thanks for your help. Mar 13 12:19:42 well usually it's quit straightforward, not sure why make toolchain did not work out for the wl500g Mar 13 12:21:09 I'v read about the exceptions-issues you mentioned so it is probably related to C++ apps.. Mar 13 12:21:32 yes, it's an issue with uclibc++ that oleg firmware was using, not sure what openwrt is using Mar 13 12:22:25 The uclibc++ from the optware feeds Mar 13 12:22:30 In this case. Mar 13 12:24:14 well, let's see how far I get with the static builds, but it would be of course nicer to have it working via optware Mar 13 12:24:48 When are you planning on releasing them_ Mar 13 12:24:50 ? Mar 13 12:25:09 not sure, so far I am working on a build env (OE based) that would allow me to build for all kinds of targets Mar 13 12:25:32 so depends on how fast I get the setup to do what I want Mar 13 12:25:39 Alright Mar 13 12:26:35 btw, is there support for album-art in id3-tags? Mar 13 12:27:26 yes Mar 13 12:27:39 but PS3 will only show them if they are not bigger than 120x120 Mar 13 12:28:30 Great. Both through taglib and id3lib? Mar 13 12:28:37 yes Mar 13 15:29:02 mandersson & Jin^eLD : openwrt-brcm24 is a different target than wl500g Mar 13 15:29:16 make openwrt-brcm24-target Mar 13 15:36:52 eno: aaah.. thanks Mar 13 15:52:22 I'm using the openwrt-brcm24 target Mar 13 15:53:48 ok, taglib indeed fails there Mar 13 15:53:50 but id3lib works Mar 13 15:54:34 eno: how would one correctly add an "if no taglib, try id3lib" dependency to the .mk ? Mar 13 15:54:39 Yes but I am having trouble building mediatomb and making it link against the right C++ lib Mar 13 15:56:43 mandersson: well, if their uclibc++ can not handle exceptions then you need to link vs libstdc++ Mar 13 15:56:47 ..so I think Mar 13 15:58:27 Yes but as I mentioned before, whoever put the package in the feed managed to build an link against uclibc++ and I am able to run it so.. Mar 13 15:59:24 are you sure it did not depend on libstdc++? Mar 13 15:59:36 maybe libstdc++ was just pulled in and you did not notice? Mar 13 16:00:30 ldd tells me it's uclibc++ for the binary in the feed and libstd++ for the binary I build Mar 13 16:01:12 And I don't even have stdlibc++ installed. Mar 13 16:01:28 The packaged in the feed named libstdc++ installs uclibc++ Mar 13 16:01:38 hm Mar 13 16:02:27 Just creating a link to uclibc++ doesn't work.. Mar 13 16:04:03 I guess I should send an e-mail to the package maintainer... Mar 13 16:04:13 http://www.deadlock.dhs.org/jin/mt/bcm/ Mar 13 16:04:25 can you try the packages I built and see what it is linked against? Mar 13 16:05:14 Sure. Hang ong Mar 13 16:06:18 Jin^eLD: taglib then id3lib logic sounds good Mar 13 16:09:57 uclibc++ Mar 13 16:10:00 Ok Mar 13 16:10:27 So my toolchain is obviously messed up somehow Mar 13 16:11:19 I'm using the "MasterMakefile" to set up the build env. Mar 13 16:19:09 eno: I am not sure how to add that logic in the .mk file Mar 13 16:20:25 mandersson: the only thing I did was : check out optware, cd trunk, make openwrt-brcm24-target, cd openwrt-brcm24, make directories, make toolchain, make mediatomb Mar 13 16:20:48 since id3lib is in PACKAGES on all platforms, we can do: Mar 13 16:21:14 ifeq (taglib, $(filter taglib, $(PACKAGES))) Mar 13 16:21:21 # use taglib Mar 13 16:21:23 else Mar 13 16:21:28 # use id3lib Mar 13 16:21:31 endif Mar 13 16:21:39 sounds cool,thanks Mar 13 16:21:52 I'll see that I make a patch then, did not catch rwhitby yet Mar 13 16:34:53 Jin: Thanks for all the help Mar 13 16:35:01 np Mar 13 16:35:05 It turned out the problem was in using the Master Makefile... Mar 13 16:35:23 Checking out the toolchain manually solved the problem. Mar 13 16:37:04 sayonara Mar 13 17:44:14 eno: when my configure tries to link with id3lib for the openwrt-brcm24 target, I get undefined reference to `std::basic_ostream and stuff like that - so something c++ related Mar 13 17:44:21 am I missing some special flag ? Mar 13 18:17:56 It only worked half-way.. Mar 13 18:18:03 I know, it failed for me too Mar 13 18:18:10 I was not able to actually link vs id3lib for some reason Mar 13 18:18:35 I managed that but there are unresolved symbols at run-time.. Mar 13 18:19:04 like what? Mar 13 18:19:11 bin/mediatomb: can't resolve symbol '_ZNSs6assignEPKcj' Mar 13 18:19:46 uhm.. Mar 13 18:20:23 Don't know if there are more since it bails there.. Mar 13 18:20:54 But it seems libc++ related.. Mar 13 20:22:34 I just ran into someting odd Mar 13 20:22:56 I went to go rebuild my slug, having replaced a dead HD Mar 13 20:24:07 but when I try to turnup disk -i /dev/sda1 it says tere's no space left on device to write linuxrc Mar 13 20:24:10 /sbin/turnup: 863: cannot create /tmp/flashdisk.1788/linuxrc.new: No space left Mar 13 21:11:06 03bzhou * r7971 10optware/trunk/make/swi-prolog.mk: swi-prolog: 5.6.51 -> 5.6.52 Mar 13 22:33:30 I learned something interesting today Mar 13 22:34:11 aparently /usr must appear on the root partition of slugosBE or the thing goes berserk Mar 13 22:35:01 Is there a method by which I can restore the /usr directory in flash or should I just reflash the box? Mar 13 23:20:20 You removed it from the flash? or the hard drive? Mar 14 00:54:40 What's the reasoning behind not building all packages for each sucessive release? Mar 14 00:54:56 Can you be more specific? Mar 14 00:55:52 U reflashed from 3.10 to 4.8 and found many of the packages I expected to be available aren't Mar 14 00:56:29 I was upgrading my callweaver box, after a drive failure, figured I might as well reflash while I was at it Mar 14 00:56:54 I'm finding libldap and tiff which weere available in 3.10 aren't in the 4.8 feeds Mar 14 00:57:10 as well as the packages I'd contributed Mar 14 00:57:54 I guess I can build them all locally, but it just struck me as odd. Mar 14 00:58:15 Is it intentional or an error? If intentional, what's the reasoning? Mar 14 01:03:39 Send an email to the list with the specific set of packages you're interested in. Basically, when a package is added to the feeds, it's there as long as it built correctly once. If the autobuilder fails on a package, it gets removed from the feeds - but the older, working package is not removed. Mar 14 01:04:21 So when the feeds are switched to a new release, only the currently-building packages end up in the feeds. Mar 14 01:05:12 Until someone requests a package, and possibly submits the necessary patches to make it work again depending on the nature of the build failure, it won't get gratuitously added to the feeds. We don't have facilities to test. :( Mar 14 01:06:20 hmm Mar 14 01:06:36 so how do we extract the autobuild error information? Mar 14 01:07:13 I bet I know why openpbx doesn't build - the distro files were moved Mar 14 01:07:26 I have a major new version of the recipe anyway Mar 14 01:07:54 but the dependency packages are what I'm nore concerned about Mar 14 01:08:23 so what it comes down to, is I now have to debug a ton of additional packages Mar 14 01:08:26 fun fun Mar 14 01:12:22 hillct: if you can build it on Ubuntu 7.10, using only the packages specified in make setup-host-ubuntu target in the Master makefile, then the autobuilder should be able to build it. Mar 14 01:12:36 if it requires new packages to be installed on the autobuilder, we can talk about that. Mar 14 01:13:00 In general, just request new OE packages on the developer list, and we'll add them if they build. Mar 14 01:13:18 K Mar 14 01:13:27 my build enviroment is Fedora Mar 14 01:14:16 if you can build it in OE there, and send the list of packages that you want that build there, then we're willing to test a build Mar 14 01:14:23 but I doubt it's a matter of build enviroment packages as these dependencies built before. Or was the build box recently upgraded? Mar 14 01:14:33 K Mar 14 01:14:41 I need to develop a full list Mar 14 01:14:49 OE is never stable, things that built before often no longer build for various reasons. Mar 14 01:15:06 I'm remembering that now... Mar 14 01:15:09 and the build box has been upgrade a number of times on the last couple of years Mar 14 01:15:33 it's been a while since I did anything with this. Once I got OpenPBX building, I left it alone for a long time Mar 14 01:15:51 it's been probably 6 months Mar 14 01:15:58 I forget exactly Mar 14 01:16:06 I'll generate the package list Mar 14 01:18:56 There's been *a lot* of disruptive changes in the build env. and OE in just 6 months. For the better, we hope, but disruptive never-the-less. Mar 14 01:19:23 aparently so... Mar 14 01:21:39 looks like there are only 2 Mar 14 01:21:46 tiff (>= 3.7.2) libldap-2.3-0 (>= 2.3.11) Mar 14 01:22:23 should I send it to the mailing list or is it good enough to mention it here? Mar 14 01:23:37 tiff seems to build Mar 14 01:24:19 openldap fails Mar 14 01:24:30 it's not in stable or untable... duno why (tiff) Mar 14 01:24:31 tiff seems to be in the packages to be built alread Mar 14 01:24:57 Packaged contents of tiffxx into /home/slug/slugos/tmp/deploy/ipk/armv5teb/libtiffxx3_3.7.2-r4_armv5teb.ipk Mar 14 01:24:58 shows up as an unsatisfied dependency Mar 14 01:25:09 oh Mar 14 01:25:14 that's what bitbake -b openembedded/packages/libtiff/tiff_3.7.2.bb does Mar 14 01:25:23 different naming convention Mar 14 01:25:29 maybe Mar 14 01:25:58 openldap fails to build Mar 14 01:26:40 if you can get openldap to build on ubuntu, then we'll put it in the feeds. Mar 14 01:26:48 K Mar 14 01:27:50 the packages with those dependencies are aparently just sub-packages of the main callweaver/openpbnx package so this may work without them Mar 14 01:27:53 still testing Mar 14 01:29:28 hmm. it appears my package was installed even though the dependency was not satisfied Mar 14 01:29:31 that'd bad Mar 14 01:31:55 yeah, the tiff issue appears to be a difference in package name. The file needed is libtiff.so.3 Mar 14 01:31:58 checking Mar 14 01:33:58 rwhitby referred to this general sort of problem as "bitrot" (I'd never heard that term before, but it's quite appropriate!) Mar 14 01:37:59 guess I'll have to deal with this later Mar 14 01:38:26 on the bright side, now I have a direction though Mar 14 01:38:29 thanks **** ENDING LOGGING AT Fri Mar 14 02:59:57 2008