00:02.14 | *** join/#oe keith_ (~keith@bh02i525f01.au.ibm.com) |
00:03.09 | *** part/#oe jmau (~jmau@pD9E5DADF.dip.t-dialin.net) |
00:05.47 | beewoolie | Is there a glibc optimizer stage I can run? Like the one that debian used (uses?) in the installer. |
00:10.17 | RP | "The LZ9JG19 and the LH58123 are exclusive ICs for a limited customer". That's me told then... |
00:10.24 | *** join/#oe Tamath (~tamath@tamath.gotadsl.co.uk) |
00:10.41 | RP | (That was the response from Sharp to my request for information) |
00:11.36 | pb_ | beewoolie: no, but it'd be very cool if you'd like to contribute one. |
00:18.41 | CosmicPenguin | the way this is acting, you would think nobody ever build openldap with gcc 3.4 before |
00:18.41 | beewoolie | pb_: well, it ought not be hard, but it may be something that has to run when the filesystem itself is built. |
00:18.49 | pb_ | yeah |
00:21.24 | *** join/#oe dkey| (~dkey@L0003P21.dipool.highway.telekom.at) |
00:24.04 | kergoth | pb_, beewoolie: glibc optimizer? what does it do? |
00:24.16 | kergoth | went home, no electricity yet |
00:24.17 | pb_ | kergoth: I think he means mklibs, the library reducer. |
00:24.18 | kergoth | heh |
00:24.20 | pb_ | heh |
00:24.20 | kergoth | ah |
00:26.24 | pb_ | This GDB was configured as "arm-linux"...Segmentation fault |
00:26.25 | pb_ | root@h3900:~# |
00:26.27 | pb_ | suck |
00:26.37 | kergoth | heh |
00:26.50 | kergoth | i saw that behavior too, makes it not very useful |
00:26.51 | beewoolie | kergoth: it removes symbols from the libraries that aren't needed by applications. |
00:26.52 | kergoth | | ../../../config/./nsinstall -R -m 444 /home/kergoth/code/build-tosa/tmp/work/arm-linux-uclibc/minimo-0.0cvs20050104-r7/mozilla/nsprpub/pr/include/md/ /home/kergoth/code/build-tosa/tmp/work/arm-linux-uclibc/minimo-0.0cvs20050104-r7/build-arm-linux-uclibc/dist/include/nspr |
00:26.56 | kergoth | | ../../../config/./nsinstall: cannot make symbolic link /home/kergoth/code/build-tosa/tmp/work/arm-linux-uclibc/minimo-0.0cvs20050104-r7/build-arm-linux-uclibc/dist/include/nspr/md: File exists |
00:27.00 | kergoth | | make[6]: *** [export] Error 1 |
00:27.20 | kergoth | heh |
00:27.29 | kergoth | applications in general, or the ones currently installed? |
00:27.44 | beewoolie | btw: pxa built OK. Thanks for the help. I'd be happy to make some edits to the sample conf to make it easier for lunk-heads like me. |
00:27.55 | kergoth | patches welcome |
00:29.18 | pb_ | drat, the gnutls people want me to fill in a copyright assignment for a tiny patch |
00:29.59 | beewoolie | the library optimizer reads the dynamic link symbols from so's and executables and removes the unused symbols from the library. The idea of 'installed' isn't really compatible with this thing. Obviously, adding a package might be a proble, |
00:30.23 | kergoth | gotcha |
00:30.31 | kergoth | so its only really useful for relatively static environments |
00:30.35 | pb_ | yeah |
00:30.42 | kergoth | that's cool though |
00:30.42 | CosmicPenguin | Still nice to have |
00:30.43 | beewoolie | Not really. |
00:30.48 | kergoth | for initrds, or what have you |
00:30.56 | kergoth | bleh, headache |
00:30.59 | CosmicPenguin | will it remove an entire library if no symbols are used? |
00:31.08 | beewoolie | It turns out that most symbols are used by the applications most people use. |
00:31.19 | beewoolie | It might make the library empty. |
00:31.33 | pb_ | CosmicPenguin: it will generate a stub .so, otherwise programs that were linked with it wouldn't be loadable. |
00:31.39 | beewoolie | In other words, once it runs, most applications will still run. |
00:31.46 | CosmicPenguin | pb_: true... |
00:31.57 | kergoth | that's pretty cool. |
00:32.08 | CosmicPenguin | pb_: back in the buildroot days I had a perl script that used ld to eliminate unused libraries |
00:32.13 | CosmicPenguin | again, only useful in static envrionments |
00:32.23 | beewoolie | It's handy for getting the bulk out of libc. ulibc, I believe, is another way to do this. |
00:32.29 | CosmicPenguin | but my primary target right now is static, so i would be very interested in such a beast |
00:32.41 | CosmicPenguin | NOTE: package openldap: build completed |
00:32.44 | CosmicPenguin | Will wonders never cease |
00:32.48 | kergoth | heh, cool |
00:32.54 | beewoolie | I'd guess that a non-static environment hasplenty of space. |
00:33.03 | kergoth | beewoolie: that isnt necessarily the case. |
00:33.10 | pb_ | CosmicPenguin: it's just a little python script. you can get it from debian-boot cvs. |
00:33.10 | kergoth | beewoolie: pdas like zaurus and ipaq for example |
00:33.20 | pb_ | might need a bit of tweaking for cross operation, but probably not much. |
00:33.27 | CosmicPenguin | Whats it called? |
00:33.30 | pb_ | mklibs.py |
00:33.31 | keturn | I have a static target too, but that sounds scary to me. I'll have to be pretty pressed for space before I resort to that. |
00:33.50 | *** join/#oe Cwiiis (~cwiiis@user-214-207-151-83.e7even.com) |
00:34.17 | beewoolie | I set ASSUMES_PROVIDED=virtual/arm-linux-gcc and I'm still getting a gcc build. I wonder if this isn't the wrong ASSUMES. |
00:34.46 | kergoth | beewoolie: its ASSUME, not ASSUMES. you're using an external toolchain i presume? |
00:34.52 | beewoolie | kergoth: well, when space is tight, especially floppy disk sized tight, something as to give. |
00:35.35 | beewoolie | Right. ASSUME. Yes. I'd like to use a prebuilt toolchain. Phil gave me this formulation. |
00:35.36 | CosmicPenguin | <PROTECTED> |
00:35.37 | CosmicPenguin | nice |
00:35.48 | kergoth | beewoolie: yes, i know. that doesnt have anything to do with your statement that a non-static environment has plenty of space. pdas generally dont have plenty of space, but do generally allow installing and uninstalling of packages. |
00:37.09 | beewoolie | The standard libc is ginormous. As we all know. IIRC, the optimizer game me a couple of hundred K back. It might even have been more, but I don't recall at the moment. |
00:40.35 | beewoolie | pb_: I think that I used the py and the shell versions both. IIRC, there was a reason why I abandoned the python version. It might have been slower than the shell script. |
00:40.56 | *** join/#oe CoreDump|hom (~mhentges@xdsl-213-196-226-123.netcologne.de) |
00:41.23 | pb_ | beewoolie: seems unlikely. the mklibs.sh one was almost always much slower. |
00:41.37 | pb_ | iirc, speed was the primary advantage of mklibs.py in the first place |
00:42.20 | Tygger-Bob | Chris.. can you do that update thing again? Cwiiis updated matchbox.. |
00:42.26 | beewoolie | pb_: well, it's been at least two years since I used it. It might have simply been a problem running the python version. |
00:42.29 | kergoth | Tygger-Bob: will do |
00:42.38 | Tygger-Bob | please, btw.. and thank you. |
00:42.45 | kergoth | neat, real actual traffic on the unionfs mailing list |
00:42.51 | kergoth | heh |
00:43.24 | beewoolie | The ASSUME_ switch seems to really not have worked. Anyone used it to coerce the use of another toolchain? |
00:43.42 | kergoth | yes |
00:44.05 | kergoth | ASSUME_PROVIDED = "virtual/${TARGET_PREFIX}gcc virtual/libc" |
00:44.12 | kergoth | that's in my build-omap-exttc/conf/local.conf |
00:44.15 | beewoolie | kergoth: I've got ASSUME_PROVIDED = "virtual/arm-linux-gcc" |
00:44.23 | kergoth | which is my tree for building packages against an external toolchain for work |
00:44.42 | beewoolie | kergoth: ah. Phil's way doesn't work. Changing it... |
00:45.01 | kergoth | having virtual/arm-linux-gcc should be fine, as long as your TARGET_ARCH is arm and your TARGET_OS is linux |
00:45.51 | kergoth | it'll build a C library, mine wont |
00:45.51 | kergoth | beyond that, either should work |
00:45.51 | beewoolie | Isn't linux the default targetOS? |
00:45.52 | kergoth | technically, the default target os is the build os |
00:45.55 | kergoth | and the default build os is read out of a call to uname |
00:46.04 | kergoth | so if you're building on cygwin, or a bsd, thatll be your default target as well |
00:46.12 | kergoth | but yes, in most cases the default would be linux |
00:46.14 | CIA-10 | 03pb 07 * r1.2865 10openembedded/packages/gpe-aerial/ (files/fix_makefile.patch gpe-aerial_0.2.12.bb): update gpe-aerial to 0.2.12 |
00:46.39 | beewoolie | kergoth: well, your method puts out the same ASSUME_ that mine did, except for the library. |
00:46.51 | kergoth | right, and i'm saying either shouldve worked |
00:46.57 | beewoolie | kergoth: right. |
00:47.38 | beewoolie | I wonder if this isn't the problem: I've got the PREFERRED_PROVIDERS with " virtual/${TARGET_PREFIX}gcc:gcc-cross" |
00:47.53 | kergoth | looks fine. |
00:47.57 | beewoolie | kergoth: that makes me think it's gonna force the use of gcc-cross |
00:48.02 | kergoth | .. |
00:48.15 | kergoth | gcc-cross is one of 3 different packages that PROVIDES virtual/${TARGET_PREFIX}gcc |
00:48.26 | kergoth | you're saying that virtual/${TARGET_PREFIX}gcc is already provided, so it doesnt have to build it at all |
00:48.28 | beewoolie | ...and doesn't it build a compiler? |
00:48.31 | kergoth | regardless of which provider you prefer. |
00:48.43 | beewoolie | hmm. We'll gcc-cross is the package being built. |
00:48.45 | kergoth | all PREFERRED_PROVIDERS does is gives you a way to exert control over multiple provides situations |
00:48.48 | kergoth | tahts all |
00:48.50 | kergoth | of course it is. |
00:48.59 | kergoth | its trying to build virtual/${TARGET_PREFIX}gcc |
00:49.06 | kergoth | it should not be if tahts listed in ASSUME_PROVIDED |
00:49.17 | kergoth | your preferred providers is just a preference. it isnt forcing anything |
00:49.25 | kergoth | and it will not subvert what assume_provided does |
00:49.33 | beewoolie | I'm assuming that if I say don't build gcc, then it won't build gcc-cross. |
00:49.42 | kergoth | ? |
00:49.46 | kergoth | you arent saying dont build gcc. |
00:49.53 | kergoth | you're saying dont build virtual/${TARGET_PREFIX}gcc |
00:49.59 | beewoolie | That's what I mean. |
00:50.17 | kergoth | if a package in the tree directly depended on gcc-cross, that would be built regardless |
00:50.21 | beewoolie | If I say don't build arm-linux-gcc, then I assume it won't build gcc-cross. |
00:50.26 | kergoth | but nothing does. everything depends on virtual/${TARGET_PREFIX}gcc |
00:50.30 | beewoolie | Well, that's what may be the case. |
00:50.32 | kergoth | that's correct |
00:50.35 | kergoth | no, that isnt the case. |
00:50.39 | kergoth | unless you changed something to do so |
00:50.42 | beewoolie | tee hee. |
00:51.01 | beewoolie | is there a command to dump the dependency tree/ |
00:51.08 | kergoth | bitbake --help |
00:51.20 | kergoth | try bitbake -v --dry-run [whateveryouwanttobuild] |
00:58.27 | beewoolie | kergoth: nothing in the dry-run |
00:59.20 | beewoolie | Hmm. could be g++, though. |
01:00.17 | CosmicPenguin | hmmm... I wonder why AC_CHECK_HEADERS is drawing a blank |
01:01.31 | kergoth | beewoolie: nope, there's no seperate provider for g++. |
01:01.43 | kergoth | beewoolie: run bbread, confirm ASSUME_PROVIDED is set the way it should be |
01:02.16 | beewoolie | kergoth: did. ASSUME_PROVIDED is what we expect. |
01:03.01 | pb_ | beewoolie: what do you mean by 'nothing in the dry-run', exactly? |
01:03.24 | beewoolie | No dependencies listed for gcc-cross. It is trying to build it, though. |
01:04.02 | pb_ | I doubt it's just trying to build it completely at random. What does it say? |
01:04.17 | beewoolie | Hmm. It looks like adding an ASSUME field for g++ eliminated the gcc-cross dependency. I'm going to see what happens when I remove it again. |
01:04.59 | kergoth | ah, there is indeed a virtual/arm-linux-g++ provides, but i didnt think anything depended on ti |
01:05.02 | kergoth | heh |
01:05.05 | kergoth | it |
01:05.09 | *** join/#oe darkschneider (~gab@81-208-36-80.fastres.net) |
01:05.45 | beewoolie | This is the first time in a LONG time that I've wanted a faster computer. |
01:05.56 | pb_ | kergoth: mm. nothing seems to. |
01:05.59 | Cwiiis | beewoolie: Know how you feel |
01:07.55 | beewoolie | Gosh darn it. It isn't doing it now. |
01:08.00 | kergoth | hehe |
01:08.10 | beewoolie | whatever |
01:08.31 | CosmicPenguin | `struct std::ios::seek_dir' has not been declared |
01:08.42 | CosmicPenguin | That has clusterfuck written all over it |
01:09.23 | CosmicPenguin | are there any C flags to make G++ 3.4 much less restrictive? |
01:10.15 | beewoolie | if the structure hasn't been declared? Is a forward sufficient? |
01:11.45 | pb_ | CosmicPenguin: -fpermissive? |
01:12.01 | beewoolie | OK. somethings not right here. I'm looking that the process table and I'm seeing that a cross compiler from the TMPDIR/cross directory is being run. |
01:12.22 | pb_ | TMPDIR/cross goes on the front of your $PATH. |
01:12.36 | pb_ | If there's a compiler in there, it will be the one that gets used. |
01:12.50 | beewoolie | But I told it to clean gcc-cross.... |
01:13.04 | pb_ | That doesn't clean the cross or staging directories. |
01:13.13 | beewoolie | Is it safe to remove the cross directory? |
01:13.38 | CosmicPenguin | Oh, shit - I know why its doing that |
01:13.40 | pb_ | If you're providing gcc and binutils externally, yeah, I guess so. |
01:19.30 | beewoolie | Is the BBFILE_PRIORITY priority higher for higher or lower numbers? docs don't say. |