[12:14] cp -p build-min/bash debian/bash-minimal/bin/bash-minimal [12:14] make: : Command not found [12:14] make: *** [binary-minimal] Error 127 [12:14] doko: ^^^ === doko_ [n=doko@dslb-084-059-077-153.pools.arcor-ip.net] has joined #ubuntu-toolchain [12:14] hrm.. maybe doko_ needs this too.. [12:14] cp -p build-min/bash debian/bash-minimal/bin/bash-minimal [12:14] make: : Command not found [12:14] make: *** [binary-minimal] Error 127 [12:15] lamont, ? [12:15] lamont__, ? [12:16] that's bash [12:17] although I must admit that it's strange [12:17] yes, i386/powerpc/amd64 did succeed ... [12:18] ok === jbailey [n=jbailey@modemcable139.249-203-24.mc.videotron.ca] has joined #ubuntu-toolchain [07:04] infinity: please requeue gcc-4.0 on i386 [07:06] Done. [07:07] hmm, maybe the same for amd64 [07:07] (Did that too at the same time) === chmj [n=chmj@dsl-146-190-110.telkomadsl.co.za] has joined #ubuntu-toolchain === fabbione [n=fabbione@port49.ds1-van.adsl.cybercity.dk] has joined #ubuntu-toolchain [01:01] doko: ping? [01:03] fabbione: pong [01:04] doko: do you happen to know if binutils is foobar on ppc? [01:04] if not, do you have a ppc handy to do a test for me? [01:05] a) no, b) sorry, no. [01:05] ok [01:05] let me explain the problem [01:05] perhaps we can reproduce it on davis [01:06] i compile libfoo with bar support [01:06] that gets linked properly [01:06] and i can see from objdump -x |grep NEEDED [01:06] that bar requires libfoo [01:07] bar pkg also ships a libbar [01:07] that's still linked properly [01:07] now [01:07] there is libbaz that B-D on libbar-dev [01:07] meh [01:07] no sorry [01:07] wrong example [01:08] doko: wipe it [01:08] let's start from scratch [01:08] with real names [01:08] i have devmapper that B-D on libselinux1-dev [01:08] devmapper produces a libdevmapper that's linked with libselinux [01:08] (verified via objdump [01:09] and up till here everything is nice and dandy [01:09] later on comes to play lvm2 [01:09] that B-D on libdevmapper [01:09] lvm2 has the option to build with or without libselinux [01:09] the first case is done for the .deb [01:09] the latter for the .udeb [01:09] now [01:10] the interesting part is that lvm2 pulls in selinux only to do a call to: [01:10] #ifdef HAVE_SELINUX [01:10] if (!set_selinux_context(lv_path, S_IFLNK)) { [01:10] stack; [01:10] return 0; [01:10] } [01:10] #endif [01:10] so if we agree.. lvm2 needs libselinux for set_selinux_context symbol [01:10] am i right? [01:10] now [01:10] the build is done without selinux [01:11] grep set_selinux_context lvm [01:11] | wc -l = 0 [01:11] so the binary doesn't have any reference to that symbol [01:11] but [01:11] objdump -x lvm | grep NEEDED [01:11] NEEDED libdevmapper.so.1.01 [01:11] NEEDED libdl.so.2 [01:11] NEEDED libc.so.6 [01:11] NEEDED libselinux.so.1 [01:11] DA DA DA [01:11] and it looks ok on other architectures? [01:11] libselinux is there [01:11] yes.. it looks ok on other arches and in Debian [01:12] the problem seems to be Ubuntu specific [01:12] i did check the build logs and stuff [01:12] i am sure -DHAVE_SELINUX is not defined [01:13] just checked binutils, no powerpc patches in ubuntu [01:13] what else could it be that causes such a thing? [01:14] it could also be that Debian has a different binutils? [01:14] how is lvm linked? [01:14] powerpc-linux-gnu-gcc -o lvm dumpconfig.o formats.o lvchange.o lvconvert.o lvcreate.o lvdisplay.o lvextend.o lvmchange.o lvmcmdline.o lvmdiskscan.o lvreduce.o lvremove.o lvrename.o lvresize.o lvscan.o polldaemon.o pvchange.o pvcreate.o pvdisplay.o pvmove.o pvremove.o pvscan.o reporter.o segtypes.o toollib.o vgcfgbackup.o vgcfgrestore.o vgchange.o vgck.o vgcreate.o vgconvert.o vgdisplay.o vgexport.o vgextend.o vgimport.o vgm [01:14] erge.o vgmknodes.o vgreduce.o vgremove.o vgrename.o vgscan.o vgsplit.o lvm.o -Wl,--export-dynamic -L../lib -L/lib -llvm -ldevmapper -ldl -rdynamic [01:14] this is the call to link lvm [01:15] is the symbol in liblvm? [01:15] grep set_selinux_context liblvm.a [01:15] fabbione@daltanius:/usr/src/lvm2-2.01.14/debian/build/build-udeb/lib$ [01:15] nope [01:15] it's used there if -DHAVE_SELINUX [01:17] brb [01:17] any of the .la files reference the library? [01:17] hmm [01:17] fabbione@daltanius:/usr/src/lvm2-2.01.14/debian/build/build-udeb$ find . -name "*.la" [01:17] fabbione@daltanius:/usr/src/lvm2-2.01.14/debian/build/build-udeb$ [01:18] libdevmapper matches.. but that's ok on the other arches [01:18] or better [01:18] it matches everywhere [01:19] but libselinux is not pulled on i386 or amd64 [01:19] i am getting workraved [01:19] brb [01:24] re [01:24] any idea? [01:26] no, currently not. [01:29] ok [01:29] we will have to ask jbailey to recheck [01:29] or actuall [01:29] there is a test i could do [01:30] trying to use a libdevmapper.a that doesn't use libselinux [01:31] and see if it actually comes from there [01:32] yes, it definetely comes from there [01:33] if i use a libdevmapper without selinux, lvm doesn't NEEDED libselinux [01:33] this is definetely something to do with binutils [01:56] fabbione: Sorry, what am I checking? [01:56] I'm just catching up on the backlog. [01:58] jbailey: read the backlog.. if there is something not clear i will explain again.. i need to take a break [02:17] re [02:18] fabbione: Without having looked it up, my best guess is that it's not a binutils issue so much as somethings reducing the symbols not being used or something. [02:18] I'll take a look a bit later, I'm still going through email. [02:19] jbailey: ok. is everything i wrote understandable? [02:19] Sort of. [02:19] the pkgs we have in the archive now are all libselinux disabled [02:19] so you will need to renable them [02:19] AFAICT, you think seliniux should be in the NEEDED set and it isn' [02:19] t on PPC. [02:19] ? [02:19] the other way around [02:19] it's NEEDED and it shouldn't [02:19] but only for the build that explicitly define --disable-selinux [02:20] Ah, okay. [02:20] note: build-udeb [02:20] Can I reproduce this on ppc? [02:20] Or sparc? [02:20] Those are my two easily-acceisble dapper machines atm. [02:20] ppc [02:20] i386 and amd64 looks ok [02:20] it seems to be a specific ppc problem [02:21] Lovely. [02:21] Always nice to use the faster machine for debugging. [02:21] =) [02:21] ehehe [02:21] And that's lvm2, right? [02:21] yes [02:21] you need to start from devmapper [02:22] enable selinux on devmapper [02:22] Hmm, silly question. [02:22] eheh ok :) [02:22] Lemme get this straight again (sorry, I'm still sleepy.. *g*) [02:22] I see devmapper has --disable-selinux. [02:22] And it's still pulling in libselinux anyway? [02:22] fabbione let's start from scratch start from here [02:23] fabbione let's start from scratch <- start from here [02:23] Or perhaps I should grab breakfast. =) [02:23] exactly [02:23] wake up first :) [02:24] a'ight. [02:59] fabbione: 'kay. I think I'm awake now. [03:00] ok [03:04] So, the ultimate goal here is to enable selinux, or not? [03:04] doko was saying to me that enabling selinux is a dapper+1 goal. [03:08] ehh, no, ssp first ... [03:09] jbailey: great that pitti isn't here ... [03:10] jbailey: ok [03:10] simple nice and dandy [03:10] selinux support should be enabled in the .debs [03:10] but NOT in the udebs [03:10] on ppc due to something fucking up [03:10] lvm gets linked with libselinux (the udeb build) [03:11] it doesn't happen on other arches [03:11] the build logs are the same [03:11] there is no HAVE_SELINUX defined [03:11] nothing in the udeb build uses symbols from libdelinux [03:11] libselinux [03:11] the only thing is libdevmapper [03:11] it uses libselinux [03:12] libdevmapper is used both for .debs and .udebs build [03:12] this still doesn't explain why an objdump on PPC shows NEEDED libselinux [03:12] and it doesn't on other arches [03:12] this makes lvm2 fails on ppc [03:14] make more sense now? [03:17] anyway i need to go offline [03:17] i might pass by later [03:18] 'k [03:18] I'll dig through and get this. [03:18] See you after your nap. =) [04:56] nice one: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=25199 [05:04] Jakub's comment: Finding a bug in 1M of assembly is really hard without knowing where exactly [05:04] to look at. [05:06] yes, but he already posted the patch [05:06] Yes. Jakub scares me. =) [05:06] I like our gcc-opt adding -mtune=pentium4 ... === mdz [n=mdz@studiocity-motorola-bsr1-70-36-194-85.vnnyca.adelphia.net] has joined #ubuntu-toolchain [06:44] jbailey: ok thanks.. i don't think i can do more debugging really.. [06:44] Yup, it's all good. [06:44] jbailey: i need to go off for the night.. my wife invited mother and sister [06:44] I haven't looked at it yet, but I'm having a slow day today. [06:45] 'kay. Sometime next week I want to grab you for a server kernel discussionwith benc. [06:45] jbailey: no problem.. i am pretty sure it's something not related to the pkg itself [06:45] You're the approver. =) [06:45] you might notice is that in Debian doesn't happen [06:45] jbailey: sure [06:45] let's plan a meeting.. [06:45] After your nap on Monday, maybe? [06:45] perhaps monday evening around 20:00 UTC would be lovely [06:45] no after the nap i have the danish exam [06:46] 20 UTC should be fine. [06:46] but i will finish early -> early dinner -> put wife to bed and i am up for * === jbailey checks with Ben. [06:46] jbailey: please coordinate with Ben [06:46] i need to go offline again now [06:46] Doing. G'night, see you around this WE. [06:46] i might pass by later.. [06:46] not sure yet :/ [06:47] it depends if wife's sister's kids will allow me ;) [08:42] doko: I think I gave back whatever package it was you wanted given back on hppa/unstable [08:49] lamont__: I tried a build of glibc cvs last night, requires a newer toolchain. [08:49] lamont__: I'll do the work on bdale's j5k box for now. [08:49] (I have root there) [08:55] 'k === mdz [n=mdz@studiocity-motorola-bsr1-70-36-194-85.vnnyca.adelphia.net] has joined #ubuntu-toolchain === mdz [n=mdz@studiocity-motorola-bsr1-70-36-194-85.vnnyca.adelphia.net] has joined #ubuntu-toolchain === mdz [n=mdz@studiocity-motorola-bsr1-70-36-194-85.vnnyca.adelphia.net] has joined #ubuntu-toolchain