00:00.27 | mjn3 | last time i checked |
00:00.28 | Qui_Gon | i've found thx |
00:01.35 | Qui_Gon | do u know if via 800 motherboard accept usb boot ? |
00:01.57 | mjn3 | no idea. sorry |
00:02.36 | Qui_Gon | thx for all and good night |
00:02.42 | Qui_Gon | c u soon |
00:03.12 | mjn3 | good luck |
00:04.36 | Qui_Gon | thx i will be back ;) |
00:05.42 | mjn3 | sigh... |
00:06.12 | mjn3 | libdl isn't being built with soft float |
00:50.59 | *** join/#uclibc ade|home (~adrian@modem-4000.gorilla.dialup.pol.co.uk) |
03:51.47 | *** join/#uclibc ZaM (~bmod@033.c.007.syd.iprimus.net.au) |
05:43.56 | *** join/#uclibc andersee (~andersee@codepoet.org) |
05:44.03 | mjn3 | andersee: hey |
05:44.42 | andersee | hey |
05:44.47 | andersee | just got home |
05:44.53 | psypete | andersee: did you ever boot a kernel on the dreamcast via serial or ethernet? |
05:45.13 | andersee | psypete: only from cd |
05:45.18 | psypete | dam |
05:45.23 | mjn3 | andersee: i checked in a couple of changes today |
05:45.59 | andersee | k |
05:46.04 | andersee | I'll try to look |
05:46.08 | mjn3 | ldso and libdl weren't being built with soft float. neither were the uClibc utils |
05:46.13 | mjn3 | i'm doing test builds now |
05:46.54 | andersee | I've been up since 5am finland time, so I'm fading fast |
05:47.00 | mjn3 | i bet |
05:47.56 | mjn3 | there's a problem with the cris build. some needed libgcc funcs are static. i'll look into that tomorrow |
05:48.10 | mjn3 | assuming all the other builds look like they are working |
05:48.25 | andersee | mjn3: very cool |
05:48.46 | andersee | mjn3: suppose yet another release would be order? |
05:48.51 | andersee | before we break something |
05:48.54 | andersee | ;-) |
05:49.02 | mjn3 | yeah.. that's what i was thinking |
05:49.18 | mjn3 | i'd like people to test the various archs |
05:50.07 | mjn3 | tomorrow i plan to put the soft float rootfs on winder |
05:50.29 | mjn3 | then i want to build perl and python and run their self-tests |
05:52.31 | mjn3 | one issue i haven't looked into is strace not building for arm |
06:03.43 | andersee | off to bed |
06:04.05 | andersee | catch you tomorrow after some sleep so I can think again |
06:04.10 | andersee | night |
06:04.10 | mjn3 | ok |
06:04.12 | mjn3 | night |
06:46.36 | *** join/#uclibc YogSothoth (~knoppix@adsl-18-27.swiftdsl.com.au) |
07:35.00 | *** join/#uclibc tahorg (~loic@perax.net1.nerim.net) |
11:13.23 | *** join/#uclibc tahorg (~loic@perax.net1.nerim.net) [NETSPLIT VICTIM] |
11:46.05 | *** join/#uclibc TheMasterMind1 (~aman@h-68-166-69-153.MCLNVA23.dynamic.covad.net) |
12:42.01 | *** join/#uclibc ade|desk (~adavey@194.200.143.253) |
13:06.57 | *** join/#uclibc sjhill (~sjhill@65.117.135.105) |
14:14.52 | *** join/#uclibc dsmith (~dsmith@mail.actron.com) |
16:14.14 | *** join/#uclibc miguelanxo (~miguel@88.Red-80-33-34.pooles.rima-tde.net) |
16:14.17 | miguelanxo | helo |
16:14.35 | miguelanxo | does sh support this lines? |
16:14.40 | miguelanxo | . file_to_source |
16:14.56 | miguelanxo | ? It looks like busybox port of ash doesnt support that |
16:17.48 | mjn3 | miguelanxo: busybox ash does suppot 'dot' to source a file |
16:18.01 | miguelanxo | ok I know that |
16:18.07 | miguelanxo | but what about plain sh? |
16:18.19 | miguelanxo | I mean, is that a feature of bash or what |
16:18.54 | mjn3 | sourcing a file with '.' is a standard bourne shell feature. it isn't a bash-ism |
16:19.12 | mjn3 | i haven't used any of the busybox shells other than ash though |
16:19.32 | mjn3 | we are talking about busybox 1.0pre*, right? |
16:19.46 | miguelanxo | just let me confirm... |
16:20.18 | miguelanxo | busybox-1.00-pre3 in fact |
16:20.47 | miguelanxo | do tou know of any other little sh which does support . source_file? |
16:21.04 | mjn3 | ash should be supporting it. let me look |
16:21.13 | miguelanxo | I need to execute hotplug (depending on bash) in busybox |
16:21.32 | miguelanxo | what a bout a uclibc version of bash, will it compile? |
16:21.53 | mjn3 | yes, you can compile bash using uClibc |
16:23.20 | miguelanxo | what should I spect as of its file size? |
16:25.11 | miguelanxo | thank mjn3 |
16:25.20 | miguelanxo | you guys just rock! |
16:25.54 | mjn3 | bash linked with uclibc is about half a meg |
16:25.58 | mjn3 | on x86 anyway |
16:26.01 | miguelanxo | aaaah |
16:26.09 | miguelanxo | that is HUGE for me |
16:26.18 | miguelanxo | stripped? |
16:26.42 | mjn3 | i just checked and busybox ash (sh) dot sourcing is working here |
16:26.43 | miguelanxo | anyway |
16:26.58 | miguelanxo | yes? |
16:27.00 | miguelanxo | which version |
16:27.51 | mjn3 | snapshot of a few days ago. but there haven't been any busybox shell-related changes in a while |
16:27.58 | mjn3 | yes, bash is stripped |
16:30.51 | miguelanxo | look at this |
16:31.01 | miguelanxo | you have hotplug? |
16:31.33 | mjn3 | yes |
16:32.03 | miguelanxo | look at /etc/hotplug/default/default.hotplug |
16:32.07 | miguelanxo | line 27 |
16:32.24 | miguelanxo | that cd /etc/hotplug |
16:32.24 | miguelanxo | . hotplug.functions |
16:32.26 | miguelanxo | just fails |
16:32.41 | miguelanxo | maybe its related to cd insead of . |
16:33.03 | miguelanxo | no no |
16:33.05 | miguelanxo | no no |
16:33.11 | miguelanxo | just forget about that |
16:33.57 | mjn3 | miguelanxo: ahh... i see |
16:34.19 | mjn3 | it is failing for a file without a specific path |
16:34.35 | miguelanxo | no really |
16:34.38 | miguelanxo | or yes? |
16:35.00 | miguelanxo | I'm just trying to make hotplug firmware agent to work using busybvox sh |
16:35.20 | mjn3 | well, ". filetosource" fails bug ". ./filetosource" works |
16:35.30 | miguelanxo | yes? |
16:35.33 | mjn3 | s/bug/but/ |
16:35.58 | miguelanxo | yesh |
16:36.08 | mjn3 | so, a bug in busybox ash |
16:36.08 | miguelanxo | it fails even using cd before |
16:36.21 | miguelanxo | pwd shows correct /etc/hotplug |
16:36.27 | miguelanxo | but . fails |
16:36.35 | mjn3 | checking the source... |
16:36.39 | miguelanxo | but using full path it uns |
16:36.50 | miguelanxo | s/uns/runs ok |
16:43.08 | miguelanxo | looks like touching here and there (mainly adding full path to . and substituting /bin/bash for /bin/sh) now firmware agent kind f works |
16:45.41 | miguelanxo | mjn3: do you knoq anything about hotplug? |
16:45.51 | mjn3 | not really. i've never looked at it |
16:47.55 | miguelanxo | # ACTION=%s [add or remove] |
16:47.55 | miguelanxo | # DEVPATH=%s [in 2.5 kernels, /sys/$DEVPATH] |
16:47.55 | miguelanxo | # FIRMWARE=%s |
16:47.55 | miguelanxo | # |
16:48.28 | miguelanxo | maybe the kernel set $ACTION $DEVPATH and $FIRMWARE befre calling hotplug? |
16:48.39 | miguelanxo | that would be just the info I know |
16:49.14 | miguelanxo | mjn3 if you find ". file" to be a bug would you report it? |
16:49.22 | mjn3 | yes |
16:49.25 | miguelanxo | ok thenks |
16:49.28 | mjn3 | actually, i'd fix it. :-) |
16:49.38 | miguelanxo | so you're a developer? |
16:49.54 | mjn3 | yes |
16:50.42 | miguelanxo | good work, dude. Maybe I could invite you some beer next time you visitr Spain |
16:50.42 | miguelanxo | ;) |
16:50.53 | mjn3 | :-) |
17:03.55 | *** join/#uclibc andersee (~andersee@codepoet.org) |
17:06.29 | mjn3 | miguelanxo: hmm... apparently, busybox ash dotcmd behavior is correct. according to the standards, if the filename doesn't contain a slash, then PATH is searched. so unless the current dir or '.' is in your path, ". filewithoutpath" will fail |
17:08.08 | mjn3 | miguelanxo: see www.opengroup.org/onlinepubs/007904975/utilities/dot.html |
17:08.12 | mjn3 | andersee: morning |
17:08.25 | andersee | mjn3: morning |
17:08.59 | mjn3 | andersee: i fired of 12 buildroot builds last night |
17:09.06 | andersee | oh my |
17:09.20 | andersee | mjn3: you fired them all huh? |
17:09.39 | andersee | mjn3: now what'll they do for work? |
17:09.41 | mjn3 | arm, i386, mips, mipsel (all with and without soft float) plus cris, powerpc, sh4, and sparc |
17:09.52 | mjn3 | andersee: fired off that is |
17:10.00 | andersee | ;-) |
17:10.16 | andersee | how'd it go? |
17:10.26 | andersee | did they all finish? |
17:11.21 | *** join/#uclibc bryanw (~xf012502@proxyle01.ext.ti.com) |
17:11.26 | mjn3 | sh4 failed in busybox build.. pivor_root requires syscall() |
17:12.22 | mjn3 | mips and mipsel builds finish, but genext2fs fails due to lack of space. probably because i'm using reiserfs and the size calculation is off |
17:13.00 | mjn3 | cris and sparc both fail in busybox build for insmod (lack of support) |
17:13.30 | mjn3 | arm build fails for strace |
17:14.01 | andersee | hmm. for genext2fs run on reiserfs, I suppose you could just set GENEXT2_SIZE=100000 or some such |
17:14.06 | miguelanxo | ok, anyway i've just created a single file hotplug-firmware-agent that just works with sh |
17:14.07 | miguelanxo | :) |
17:15.05 | mjn3 | andersee: yeah... that just needs some tweaking |
17:16.03 | mjn3 | there are a couple of things i need to look at this morning buildroot-wise |
17:17.07 | andersee | mjn3: insmod support for cris and sparc may need some attention. Or may not be possible. Guess that needs some checking. If not possible though, then we need to make sure the stubs within busybox are used. |
17:17.22 | andersee | strace on arm concerns me though |
17:17.33 | andersee | that is one we _do_ want to be certain is working |
17:17.57 | andersee | mjn3: was that arm w or w/o fpu? |
17:18.05 | mjn3 | both |
17:18.49 | mjn3 | invalid use of undefined type `struct pt_regs' |
17:19.06 | mjn3 | in syscall.c |
17:20.04 | andersee | hmmm |
17:20.10 | andersee | thats bad |
17:20.27 | andersee | I've got arm w/o fpu built |
17:20.34 | andersee | lemme take a look |
17:23.55 | *** join/#uclibc DaveM_ (~david@h24-207-7-221.dlt.dccnet.com) |
17:24.23 | mjn3 | andersee: i just turned off pivot root and insmod and am continuing the sparc, cris, and sh4 builds |
17:27.58 | mjn3 | andersee: cris build fails. undefined refs to __Div, __Umod, __Mod, and __Udiv, which are all present in libgcc but as statics |
17:28.17 | mjn3 | s/libgcc/libccc_s.so/ |
17:28.57 | andersee | hmm |
17:30.03 | mjn3 | that should be easy enough to patch in gcc |
17:31.00 | mjn3 | andersee: sparc build failed for openssl.. no libdl |
17:31.13 | mjn3 | hmm... actually, no shared libs at all |
17:31.51 | andersee | mjn3: oh, right |
17:32.02 | andersee | mjn3: I never even started that |
17:32.54 | andersee | mjn3: we have some basic support that was there prior to our adopting our current ldso codebase, but I've never even tried to compile the sparc stuff |
17:33.18 | andersee | mjn3: I think I also forcably disabled shared libs for sparc |
17:33.26 | andersee | since I knew it would be totally broken |
17:33.33 | mjn3 | yeah, since nothing got built under ldso |
17:33.55 | andersee | mjn3: I had no problem building strace for arm |
17:33.59 | mjn3 | hmm |
17:33.59 | andersee | lemme blow away what I have and try again |
17:34.14 | andersee | could be due to an older version of something |
17:34.59 | mjn3 | which version did you build? 4.5? |
17:36.54 | andersee | yep |
17:36.59 | mjn3 | hmm |
17:37.57 | mjn3 | cool |
17:38.00 | andersee | using today's uClibc snapshot |
17:39.26 | mjn3 | andersee: somebody reported a problem a while back re gawk. something about uClibc's shared loader making requirements re the names of shared objects |
17:39.37 | ambassador | what builds root_fs_mimpsel? |
17:39.45 | ambassador | and is that an ext2 fs? |
17:39.54 | andersee | ambassador: buildroot does it |
17:39.58 | andersee | ambassador: and yes |
17:40.07 | ambassador | how does buildroot do it? |
17:40.19 | ambassador | i.e. how does it create the ext2fs? |
17:40.20 | mjn3 | ambassador: just set ARCH:=mipsel |
17:40.33 | mjn3 | in the buildroot Makefile |
17:40.40 | andersee | mjn3: gawk? perhaps prior to my ripping out the awk generated crti.o and crtn.o? |
17:40.40 | ambassador | I have it, I just want to know how it happened :-) |
17:41.00 | andersee | ambassador: little pixies sprinkle magic dust |
17:41.08 | ambassador | andersee: ahh... that explains it :-) |
17:41.13 | mjn3 | andersee: no. something about building gawk. looking in the archives |
17:41.14 | andersee | ambassador: and root_fs magically appears |
17:41.18 | andersee | mjn3: k |
17:41.26 | ambassador | so where do I get pixies? |
17:41.30 | ambassador | www.pixies.org? |
17:42.17 | dsmith | ambassador: see make/ext2root.mk |
17:42.26 | ambassador | dsmith: yeah, I saw that |
17:42.33 | ambassador | but it didn't build it |
17:42.49 | ambassador | oh... |
17:42.51 | ambassador | genext2 |
17:43.02 | ambassador | there it is |
17:43.06 | andersee | ambassador: yeah, that |
17:43.18 | ambassador | wait you said it was pixies! :-) |
17:43.25 | andersee | ambassador: it takes a directory of stuff and makes an ext2 fs from it |
17:43.27 | ambassador | how can I ever believe you again! :-) |
17:43.38 | andersee | ambassador: I guess we need to get a divorce now |
17:43.54 | ambassador | and I suppose you'll need a "pixie" lawyer! |
17:44.11 | andersee | ambassador: the only major problem with genext2fs is it doesn't handle hard links |
17:44.28 | ambassador | andersee: I'm cool w/that - hard links just confuse me anyway. |
17:45.04 | andersee | ambassador: i.e. hard link file A and B on the host system, run genext2fs, and the generated fs will have 2 copies, rather than a single hard linked copy |
17:45.20 | andersee | ambassador: I've been intending it fix that sometime |
17:45.23 | ambassador | yeah, yuck |
17:45.32 | ambassador | but in practice I guess it's no problem |
17:45.37 | andersee | ambassador: on my todo list somewhere |
17:45.49 | andersee | ambassador: but not exactly close to the top.... |
17:46.30 | mjn3 | andersee: found it. http://uclibc.org/lists/uclibc/2003-November/007537.html |
17:47.25 | mjn3 | i was misremembering. it wasn't the shared lib loader |
17:47.55 | *** join/#uclibc yashi_ (~yashi@dns1.atmark-techno.com) |
17:48.38 | ambassador | buildroot provide inetd/telnet |
17:48.38 | ambassador | ? |
17:48.41 | ambassador | telnetd |
17:49.16 | andersee | mjn3: hmmm |
17:50.03 | mjn3 | ambassador: you can configure busybox to provide them |
17:51.23 | andersee | ambassador: using the busybox version is also nice since you won't need to use inetd |
17:51.43 | ambassador | woohoo! I didn't know busybox had |
17:51.46 | ambassador | very cool |
17:52.01 | ambassador | that is pretty new I think must be |
17:52.42 | mjn3 | ambassador: you might want to do a 'make menuconfig' in busybox just to browse what's available |
17:53.38 | mjn3 | andersee: Gdb does not support native target sh4-unknown-linux-gnu |
17:54.08 | mjn3 | otherwise, the sh4 build went fine |
17:54.16 | andersee | mjn3: I think thats one we will not be able to fix |
17:54.25 | mjn3 | once i modified the busybox config |
17:54.36 | andersee | mjn3: I know I'm not feel especially motivated to write the gdb support |
17:54.38 | andersee | :-) |
17:54.47 | andersee | s/feel/feeling/ |
17:54.48 | mjn3 | andersee: one of the things i want to do is add some comments to the buildroot makefile |
17:55.23 | andersee | i.e. "Doesn't work for sh4 due to lack of upstream support" |
17:56.41 | mjn3 | yep |
17:59.03 | mjn3 | hmm... forgot to try m68k. starting a build now |
18:02.16 | dsmith | Did you guys see my patches to get tinyx to build under mipsel? |
18:03.04 | mjn3 | dsmith: i saw them. but one line was hard-coded for mips. that's why i didn't apply the patch |
18:03.37 | dsmith | mjn3: Yeah. Not sure what's the best way to handle that. |
18:03.46 | ade|desk | do you use xfree86 4.3 or the xc-01xxxxx thingy still ? |
18:04.15 | dsmith | ade|desk: I'm using the xc-xxxx stuff. I haven't gotten it running yet. |
18:04.34 | dsmith | Learning about initrd's and things. |
18:05.32 | ambassador | so I can locally modify sources/busybox.Config.h? |
18:05.39 | ambassador | is that the idea? |
18:05.45 | andersee | ambassador: yep |
18:05.53 | ambassador | what about locally patching busybox? |
18:06.02 | andersee | ambassador: that works too |
18:06.34 | ambassador | to do that I adjust make/busybox.mk? |
18:07.37 | andersee | ambassador: yup |
18:07.44 | ambassador | very cool, thank you |
18:08.24 | andersee | ambassador: you would then of course, blow away build_<arch>/busybox/ and then just run 'make' |
18:08.25 | mjn3 | ambassador: busybox.Config.h is for the older stable version |
18:08.42 | mjn3 | ambassador: busybox.config is for the 1.0pre |
18:09.02 | dsmith | TO get uclib to quit asking about mmu, I added "echo 'UCLIBC_HAS_MMU=y' >> $(UCLIBC_DIR)/.config" to make/uclibc.mk |
18:09.39 | dsmith | To the $(UCLIBC_DIR)/.configured target |
18:09.50 | dsmith | Is there a better way? |
18:09.55 | ade|desk | right home time, cya folks |
18:10.07 | *** part/#uclibc ade|desk (~adavey@194.200.143.253) |
18:12.44 | mjn3 | dsmith: set it in buildroot/sources/uClibc.config |
18:13.25 | mjn3 | should probably be the default in buildroot anyway, since buildroot pretty much assumes shared libs |
18:17.11 | andersee | mjn3: agreed |
18:17.22 | andersee | heh |
18:17.24 | andersee | http://www.brentsweather.com/ |
18:17.43 | mjn3 | andersee: hmm... i tried an m68k buildroot build. it doesn't build shared lib support either |
18:17.52 | andersee | complete with a current pic of the mountain behind my house |
18:17.59 | andersee | mjn3: right |
18:18.39 | mjn3 | so ldso/m68k is broken? |
18:18.40 | andersee | mjn3: HAVE_NO_SHARED in extra/Configs/Config.m68k |
18:18.44 | andersee | mjn3: yep |
18:18.49 | dsmith | andersee: Cool. |
18:19.09 | andersee | mjn3: I forcably disabled it since I have maintained it |
18:19.53 | andersee | err |
18:20.02 | andersee | have _not_ maintained it |
18:20.07 | mjn3 | andersee: oh... someone popped in the other day and said to tell you that he had succeeded in getting nptl to do create and join |
18:20.16 | andersee | mjn3: cool! |
18:20.33 | mjn3 | andersee: i automaticly inserted the missing negative |
18:20.41 | andersee | mjn3: there was a guy that said he was working on getting nptl working with uClibc |
18:21.06 | andersee | mjn3: nice to hear he is actaully working on it and making progress |
18:22.06 | andersee | mjn3: arm build finished sucessfully |
18:22.12 | andersee | mjn3: including strace |
18:22.17 | mjn3 | i wonder why i'm seeing a problem |
18:22.45 | andersee | mjn3: perhaps you need to order more pixie dust? |
18:24.16 | mjn3 | andersee: nope.. still dies. what kernel headers are you using? |
18:25.26 | andersee | TARGETS+=kernel-headers |
18:25.26 | mjn3 | oops.. didn't blow away the unpacked code |
18:25.50 | miguelanxo | mjn3, just about now I have tested the uDisttribution, and my uFirmware Agent just works ok! |
18:25.52 | miguelanxo | ;) |
18:26.09 | miguelanxo | thanks for your help |
18:26.20 | mjn3 | miguelanxo: you're welcome |
18:26.42 | mjn3 | andersee: it still dies. do you have any strace patches that you haven't committed? |
18:26.49 | andersee | mjn3: everything is stock buildroot except for enabling some more targets |
18:26.54 | andersee | mjn3: nope |
18:27.29 | andersee | I have a minor change to sources/hotplug.patch I've not committed, but I didn't build that |
18:28.15 | andersee | I have make/customize.mk setup to do a build for my little wireless elan router, but I didn't enable that |
18:28.23 | andersee | other than that, stock... |
18:28.44 | mjn3 | i'm still getting struct pt_regs undefined type errors |
18:29.52 | *** part/#uclibc miguelanxo (~miguel@88.Red-80-33-34.pooles.rima-tde.net) |
18:30.48 | dsmith | mjn3: So you never do a make clean, you just rm the build dirs? |
18:31.21 | mjn3 | dsmith: actually, i generally move them so i have a base of comparision if the next build breaks |
18:31.41 | dsmith | mjn3: ok. |
18:47.32 | mjn3 | andersee: what are your proc and arch symlinks for arm in the kernel headers? |
18:49.03 | andersee | mjn3: I just use what I have in /usr/src/linux/include/asm/ on winder |
18:49.27 | andersee | $ ls -l arch proc |
18:49.27 | andersee | lrwxrwxrwx 1 andersen src 12 Aug 4 19:31 arch -> arch-ebsa285/ |
18:49.27 | andersee | lrwxrwxrwx 1 andersen src 9 Aug 4 19:31 proc -> proc-armv/ |
18:49.34 | mjn3 | ok |
18:54.01 | mjn3 | ambassador: what are you trying to patch? |
18:54.06 | ambassador | declarative programming is not my forte |
18:54.35 | ambassador | I have some local patches to busybox |
18:54.47 | ambassador | like a mtdcp that insulates nand/nor differences in mtd |
18:54.54 | ambassador | and an mtderase to erase flash |
18:55.27 | andersee | ambassador: BUSYBOX_PATCH:=$(SOURCE_DIR)/busybox.patch |
18:55.37 | andersee | ambassador: cat $(BUSYBOX_PATCH) | patch -p1 -d $(BUSYBOX_DIR) |
18:55.57 | andersee | add the first line after the first "endif" |
18:56.26 | andersee | add the second line after the call to BUSYBOX_UNZIP to untar busybx |
18:56.51 | ambassador | I'm thinking that might double patch if make rerun? |
18:57.28 | mjn3 | andersee: look at how patching is done in gcc_3.3.mk |
18:57.41 | ambassador | I saw some .patched fussiness in the valgrind .mk |
18:57.45 | mjn3 | urr... ambassador that is |
19:02.14 | ambassador | is busybox ifupdown like the debian one? uses /etc/network/interfaces ? |
19:03.56 | andersee | ambassador: yes |
19:03.57 | ambassador | busybox has gone crazy! |
19:04.14 | ambassador | when did you guys like add *every app in the world* to it? |
19:04.59 | mjn3 | ambassador: while you were sleeping. it was the pixies |
19:05.22 | mjn3 | helped by the shoemaker's elves |
19:05.31 | ambassador | you saw Elf right? |
19:05.46 | ambassador | I laughed *out loud* and *a lot* |
19:06.52 | mjn3 | wow... somebody has time to take in a movie |
19:07.26 | ambassador | why busybox.Config.h *and* busybox.config? |
19:07.57 | mjn3 | the config system changed between the old stable and the newer version |
19:08.03 | ambassador | ok, that was my guess |
19:08.16 | ambassador | so for snapshot use I'll adjust busybox.config |
19:08.23 | mjn3 | yeah |
19:08.41 | ambassador | so when will busybox have an x server? :-) |
19:09.11 | mjn3 | someone actually suggested that on irc this weekend |
19:09.13 | dsmith | ambassador: tinnyx is there, but it's broken. |
19:09.16 | ambassador | ha! |
19:09.25 | ambassador | dsmith: no... |
19:09.35 | ambassador | dsmith: tinyx in busybox? |
19:09.38 | kergoth | heh, he said busybox not buildroot |
19:09.42 | dsmith | Ah. |
19:09.48 | dsmith | "Never Mind" |
19:10.38 | ambassador | busybox is like the blob :-) |
19:10.45 | ambassador | it just keeps consuming other apps :-) |
19:11.39 | ambassador | mjn3: I think there is some rule about that :-) |
19:11.56 | dsmith | Now I should have thought of that. |
19:12.02 | ambassador | I picture Escher's hand-drawing-hand |
19:29.12 | mjn3 | andersee: well, strace 4.4.94 builds fine for me on arm. but no other 4.4.9* builds, nor does 4.5 or 4.5.1 |
19:30.16 | *** join/#uclibc TheMasterMind1 (~aman@h-68-166-69-153.MCLNVA23.dynamic.covad.net) [NETSPLIT VICTIM] |
19:33.03 | andersee | mjn3: odd |
19:33.11 | mjn3 | yeah.. doing a diff now |
19:33.20 | andersee | mjn3: strace 4.5 for arm built fine for me |
19:33.34 | andersee | what was the exact error again? |
19:34.00 | mjn3 | struct pt_regs wasn't being defined. asm/ptrace.h was never included |
19:35.39 | andersee | mjn3: is asm/ptrace.h even present in staging dir? |
19:35.52 | andersee | mjn3: or is asm perhaps a broken symlink? |
19:37.18 | mjn3 | no.. staging_dir/include/asm/ptrace.h is fine. so is asm/proc/ptrace.h which it includes |
19:37.36 | andersee | how strange |
19:37.51 | mjn3 | yeah |
19:38.08 | andersee | dunno man |
19:38.12 | andersee | back shortly |
19:38.17 | andersee | gotta run to the store |
19:38.36 | mjn3 | ok.. i'll be away myself for a bit |
20:18.44 | ambassador | my root/lib/ isn't stripped |
20:18.59 | ambassador | $ file build_mipsel/root/lib/libuClibc-0.9.23.so |
20:19.00 | ambassador | build_mipsel/root/lib/libuClibc-0.9.23.so: ELF 32-bit LSB MIPS-I shared object, MIPS, version 1 (SYSV), not stripped |
20:19.12 | ambassador | so it's kinda big and so is rootfs |
20:19.31 | ambassador | t$ ls -l build_mipsel/root/lib/libuClibc-0.9.23.so |
20:19.31 | ambassador | -rw-r--r-- 1 brad brad 34637377 Nov 24 11:42 build_mipsel/root/lib/libuClibc-0.9.23.so |
20:19.45 | ambassador | (eek) |
20:20.13 | ambassador | should buildroot strip (almost) everything? |
20:23.51 | ambassador | aha... |
20:23.52 | ambassador | ext2root: genext2fs |
20:23.52 | ambassador | <PROTECTED> |
20:51.59 | psypete | anyone awake that can tell me the mknod line for /dev/nfs ? the "symbolic name" doesn't seem to be sticking... |
20:54.23 | *** join/#uclibc Qui_Gon (~fox@lns-vlq-7-82-65-221-75.adsl.proxad.net) |
20:55.19 | DaveM_ | psypete: there is no /dev/nfs |
21:05.03 | psypete | ok |
21:05.44 | Qui_Gon | hello |
21:08.12 | psypete | hello |
22:12.58 | ambassador | hrm... |
22:13.04 | *** join/#uclibc sjhill_gone (~sjhill@65.117.135.105) |
22:13.28 | mjn3 | ambassador: hrm?? |
22:14.47 | ambassador | I thought we had -Os for ldso mips |
22:14.58 | ambassador | but it's still -O0 in the snapshot |
22:16.09 | mjn3 | ambassador: do you have ld debugging turned on? |
22:16.22 | ambassador | oh... no |
22:16.28 | mjn3 | secificly SUPPORT_LD_DEBUG |
22:16.43 | mjn3 | urr.. specificly |
22:17.53 | ambassador | so snapshot is still broken |
22:18.00 | ambassador | for mips |
22:18.03 | ambassador | (el) |
22:18.40 | mjn3 | ambassador: what is the date of your snapshot |
22:19.21 | ambassador | $ ls -l sources/dl/uClibc-snapshot.tar.bz2 |
22:19.21 | ambassador | -rw-r--r-- 1 brad brad 1601647 Nov 24 02:16 sources/dl/uClibc-snapshot.tar.bz2 |
22:23.54 | mjn3 | ambassador: in uClibc/ldso/ldso/Makefile, is the ifeq (,powerpc) test insize the ifeq ($(SUPPORT_LD_DEBUG block? |
22:24.44 | ambassador | yes |
22:24.46 | ambassador | it's inside |
22:24.49 | ambassador | so... |
22:24.55 | ambassador | I must have it on |
22:24.57 | ambassador | :-P |
22:24.58 | ambassador | sorry |
22:25.33 | ambassador | I'm just not thinking |
22:25.35 | ambassador | you were right |
22:25.51 | ambassador | I'll test again from scratch soon and confirm that it's all well |
22:26.18 | mjn3 | well, i did a bunch of builds last night |
22:26.44 | mjn3 | <PROTECTED> |
22:26.56 | ambassador | yeah, it builds fine, then ld.so segfaults |
22:27.10 | mjn3 | still? |
22:27.27 | ambassador | no :-) |
22:27.36 | ambassador | well, -Os fixes the segfault |
22:27.40 | mjn3 | ah |
22:27.47 | ambassador | but I had SUPPORT_LD_DEBUG so it segfaulted :-P |
22:27.54 | ambassador | I think we must fix that Makefile |
22:28.08 | ambassador | ldso/ldso/Makefile |
22:28.43 | mjn3 | i told erik it was broken, and he moved the cflag overides into the debugging block |
22:28.46 | ambassador | and put a comment "MIPS requires inlines" |
22:29.01 | ambassador | yeah, but it's broke with debug |
22:29.16 | mjn3 | actually, i'd imagine powerpc and mips probably aren't alone in this |
22:29.53 | ambassador | could be... |
22:29.59 | ambassador | if _dl_memmove isn't inlined |
22:30.06 | mjn3 | i really wonder if it is inlines... i didn't think Os turned on inlining |
22:30.21 | ambassador | then anything that requires special setup before function calls would be broken |
22:31.15 | mjn3 | well, it would be nice to know what exactly was failing. and most people aren't going to turn on ld debugging |
22:31.48 | mjn3 | one problem was that it was keyed on the overall debugging build option before |
22:31.54 | *** join/#uclibc bug1 (~bug1@c211-28-81-139.sunsh3.vic.optusnet.com.au) |
22:32.30 | mjn3 | bug1: hey |
22:33.03 | ambassador | still I think a) it's ok to debug with -Os, and b) if it's completely broken why send people on wild goose chase when debugging ldso (like I went on) |
22:33.09 | bug1 | mjn3: hi |
22:33.48 | mjn3 | ambassador: i agree... but erik dislikes debugging with optimizations turned on |
22:34.15 | ambassador | me too, but he accepts it for powerpc, and he has no mips hw :-) |
22:34.28 | ambassador | btw, it looks like on mips libgcc_s isn't making it to root/lib |
22:34.55 | mjn3 | g1: someone was in earlier having problems with busybox ash and hotplug scripts. turned out that hotplug uses the dot command to source some files and it is failing |
22:35.37 | mjn3 | bug1: the problem is that the source'd files are not in the PATH (as required by susv3) and so it isn't finding them |
22:36.05 | mjn3 | bug1: just giving you a heads up. busybox ash behavior is correct. but hotplug is assuming a bash-ism |
22:36.38 | mjn3 | ambassador: re the ldso stuff, i'll talk to erik about it tonight |
22:36.47 | ambassador | mjn3: k |
22:36.49 | mjn3 | ambassador: i'll look into the libgcc_s problem on mips |
22:36.56 | ambassador | mjn3: :-) |
22:39.21 | bug1 | mjn3: ok, thanks |
22:40.37 | mjn3 | ambassador: do you mean the build_<arch>/root/lib or the rootfs image lib? |
22:42.00 | ambassador | build_mipsel/root/lib/ |
22:42.13 | ambassador | and then of course it dooesn't make it into root_fs |
22:42.20 | mjn3 | it is there for me |
22:42.26 | ambassador | oh. |
22:42.30 | ambassador | ok... |
22:42.40 | ambassador | I will start from scratch and try again |
22:42.50 | ambassador | maybe I messed something up |
22:43.09 | mjn3 | now, i did build gcc for the target as well. but the commands to copy them to root/lib appear to be in the stage to gcc |
22:44.47 | mjn3 | ambassador: i've got an i386 build going and it hasn't gotten to the gcc target build yet. but libgcc_s is in root/lib |
22:45.48 | ambassador | ok |
22:59.09 | *** join/#uclibc miguelanxo (~miguel@22.Red-80-33-35.pooles.rima-tde.net) |
22:59.56 | miguelanxo | mjn3: I'm now connected using uclibc distro |
23:00.16 | mjn3 | miguelanxo: cool |
23:01.10 | miguelanxo | finally, the mini firmware agent was what I needed |
23:01.37 | mjn3 | miguelanxo: i mentioned the hotplug problem to bug1, as he is more involved with busybox |
23:02.10 | kergoth | oh, the ". cant load files without a path" thing? . ./blah works, . blah doesnt |
23:02.20 | kergoth | had to patch my hotplug too to be happy with that |
23:03.29 | mjn3 | kergoth: yeah. that should be sent upstream |
23:03.50 | kergoth | yeah, no reason to not specify the path.. |
23:03.57 | mjn3 | kergoth: of course, they may have already been assimilated by the 'Borg Assimilastion SHell' |
23:04.04 | kergoth | hehe |
23:04.16 | kergoth | there are many bashisms in those scripts |
23:04.23 | mjn3 | no doubt |
23:04.37 | kergoth | type -p, $(()), fun stuff |
23:04.43 | kergoth | heh |
23:05.35 | kergoth | every time i see someone use $(()) instead of expr i just want to hit something :) |
23:05.50 | mjn3 | well, busybox ash supports it now |
23:05.56 | kergoth | oh really? thats cool |
23:05.58 | mjn3 | shell arithmetic that is |
23:06.18 | mjn3 | i think it might be a config option. but it has been there for quite some time |
23:06.19 | ambassador | ew... |
23:06.20 | ambassador | $ ls -l ../buildroot-old/toolchain_build_mipsel/linux/include/asm |
23:06.20 | ambassador | lrwxrwxrwx 1 brad brad 10 Nov 24 11:29 ../buildroot-old/toolchain_build_mipsel/linux/include/asm -> asm-mipsel |
23:06.36 | ambassador | no such thing as "asm-mips" |
23:06.43 | ambassador | both mips and mipsel use "asm-mips" |
23:06.53 | mjn3 | ambassador: ok... i'll fix it |
23:07.16 | miguelanxo | I've actually talked to hotplug developer Marcel holtmann, and he suggested the use of "diethotplug" package instead of hotplug |
23:07.45 | miguelanxo | in systems that doesn't have bash |
23:08.06 | ambassador | make/kernel-headers.mk looks right though |
23:08.23 | kergoth | diethotplug is nice, but often not sufficient. the missing extensibility is annoying, and it'd be nice to see diethotplug support the new /etc/hotplug.d/ scripts |
23:08.35 | ambassador | unless $(NOMMU) contains "el" |
23:09.07 | ambassador | oh... |
23:09.27 | ambassador | looks like $(TARGET_ARCH) instead of "$(ARCH)" ? |
23:09.59 | mjn3 | ah |
23:11.01 | ambassador | is that it you think? |
23:11.34 | mjn3 | could be. i'm looking at the make logs to see what was actually done |
23:11.47 | ambassador | changing it to $ARCH fixes it for me |
23:13.59 | ambassador | that has @if so cannot see in log? |
23:14.00 | mjn3 | ambassador: yep. that should be it |
23:17.21 | mjn3 | ambassador: although, i don't think it really makes a difference since those are only used by uClibc, which fixes the links anyway. was it actually causing a problem? or just something you noticed |
23:18.46 | ambassador | something I noticed |
23:18.56 | ambassador | I don't know if it caused a problem or not |
23:19.15 | ambassador | this was in tooolchain_build_mipsel |
23:19.27 | ambassador | so that's used by gcc |
23:19.34 | ambassador | but apparently the asm isn't used at all |
23:19.50 | ambassador | (used by gcc even static build too I mean) |
23:20.03 | mjn3 | actually, the toolchain build uses the uClibc headers. erik just put the kernel headers there |
23:20.31 | kergoth | hmm thats odd, my gcc 2nd pass cant locate ld-linux.so to link against, yet i was able to build 1st pass and glibc just fine with teh same configure options and it looks like its in the right place. odd. |
23:20.31 | ambassador | ok, so don't need them in toolchain_ at all? |
23:20.31 | mjn3 | or it should anyway |
23:20.45 | mjn3 | well, they need to be unpacked somewhere |
23:20.52 | ambassador | oh |
23:21.05 | ambassador | they eventually end up in staging_dir |
23:21.06 | mjn3 | uclibc makes a copy and hacks the symlinks |
23:21.27 | ambassador | aha, uClibc copes from toolchain_ to build_/staging_dir and fixes the symlink? |
23:21.39 | mjn3 | right |
23:21.45 | ambassador | gotcha |
23:21.53 | ambassador | well, I found a latent bug then :-) |
23:21.57 | mjn3 | heh |
23:22.10 | ambassador | maybe it would bite someone someday :-) |
23:22.23 | mjn3 | could be. i'm going to check in the fix |
23:23.37 | ambassador | actually it looks like uClibc doesn't put a symlink in toolchain_, but copies the right asm-* right to asm/ |
23:24.26 | mjn3 | right |
23:24.42 | ambassador | where is that copy? |
23:25.06 | mjn3 | uClibc/Makefile... part of the headers target i think |
23:26.39 | ambassador | too deep mystery for me to find :-) |
23:26.50 | ambassador | but it appears to work so I'll not plumb it |
23:28.47 | mjn3 | ambassador: uClibc/Makefile headers target invokes uClibc/extra/scripts/fix_includes.sh |
23:28.54 | ambassador | oh, thx |
23:29.05 | mjn3 | ambassador: then a miracle occurs... |
23:29.13 | mjn3 | :-) |
23:29.23 | ambassador | this is like glibc fix-includes I guess |
23:29.56 | ambassador | (I still don't see any "cp" in fix-includes) |
23:30.33 | mjn3 | it does symlinks in uClibc/include. the actual install does the copying |
23:31.10 | ambassador | wow :-) |
23:39.37 | bug1 | hmm, im having problems extracting binutils with bb tar |
23:40.50 | bug1 | yea, looks like a serious tar bug |
23:41.29 | bug1 | ahh i bet i know what it is too |
23:41.43 | bug1 | 0 length files cause it to read to eof |
23:42.08 | bug1 | from when i merged it with copyfd |
23:50.10 | bug1 | fixed |
23:52.42 | DaveM_ | Is there a reason that libpthread is completely stripped? |
23:52.49 | DaveM_ | s/is/isn't/ |
23:53.02 | ambassador | DaveM_: in root/lib/ ? |
23:53.39 | ambassador | no libs are stripped in root/lib/ for me unless I modify ext2root.mk |
23:53.50 | DaveM_ | I don't use buildroot, so I'm not sure on your directory naming, I just looked at the output of uClibc installs |
23:53.56 | ambassador | oh |
23:54.35 | mjn3 | bug? |
23:55.32 | DaveM_ | mjn3: ahhh... are you asking me? or are you trying to get bug1's attention? |
23:55.41 | mjn3 | asking |
23:56.19 | DaveM_ | well not stripping symbols isn't really a bug since everything still works, but I was wondering if there was a reason behind this |
23:57.24 | DaveM_ | libpthread on Arm changes from 82362 bytes to 67828, so it isn't a huge deal, but still measureable |