00:56.20 | *** join/#uclibc wrobbie (n=rob@cm74.kappa84.maxonline.com.sg) |
03:11.45 | *** join/#uclibc wrobbie (n=rob@cm74.kappa84.maxonline.com.sg) |
03:12.10 | *** join/#uclibc whirm (n=whirm@80.174.190.172) [NETSPLIT VICTIM] |
04:11.50 | *** join/#uclibc wrobbie (n=rob@cm74.kappa84.maxonline.com.sg) |
07:09.14 | Xride | morning all |
07:29.08 | *** join/#uclibc blindvt_ (n=bf@62.47.159.19) |
08:38.01 | *** join/#uclibc smithj_ (n=smithj@gentoo/developer/smithj) |
08:43.32 | SpanKY | http://uclibc.org:8010/ |
08:43.36 | SpanKY | slowly getting more online ;) |
09:32.14 | CIA-10 | 03aldot * r17903 10buildroot/package/Config.in: - busybox provides gzip |
09:34.13 | CIA-10 | 03aldot * r17904 10buildroot/toolchain/dependencies/dependencies.sh: - mention the package that usually provides makeinfo |
10:10.02 | *** join/#uclibc alex_ (i=alex@dnb.nu) |
12:06.16 | alex_ | hmm.. buildroot page says to come here but topic says go to #elinux or #edev :) |
12:16.42 | blindvt` | alex_, depends on the question, really |
12:18.41 | alex_ | toolchain related in here and package related in #e*? :) |
12:19.17 | blindvt` | alex_, for current, non "external" toolchains, i may be able to help |
12:20.01 | blindvt` | anyone who uses an external toolchain should go to bug sjhill or the guy mentioned in svn log -r17800 |
12:20.01 | alex_ | ok, but.. i don't have a question :) |
12:20.18 | blindvt` | alex_, even better ;) |
12:20.45 | alex_ | i'm haxxin some in the target/iso9660/iso9660.mk atm :) |
12:21.03 | alex_ | gonna add an option like the customize package |
12:21.15 | alex_ | atleast, that's my plan :) |
12:23.19 | blindvt` | alex_, i don't use iso9660 and ignore the customize package as well as that local.mk approach ;) |
12:26.23 | blindvt` | and that external toolchain thing is annoying me to no end |
12:29.03 | alex_ | the commit message annoyed me :) |
12:29.59 | alex_ | "Support building using an external toolchain. Questions to the mailing list and all other comments to <biteme@devnull.com>." |
12:31.08 | alex_ | but i guess it's gonna be pretty sweet once it's sorted out |
12:31.26 | blindvt` | i couldn't care less |
12:31.39 | alex_ | not that i'll ever use it |
12:32.20 | alex_ | blindvt`: do you have commit access? |
12:34.22 | blindvt` | i'll be off for 20 days starting monday, btw, vapier. So me ignoring stuff may not be 100% deliberate |
12:34.31 | blindvt` | alex_, why do you ask? |
12:34.36 | alex_ | just curious |
12:34.45 | blindvt` | alex_, i do |
12:39.12 | blindvt` | that slattach proposal must be a joke. A bad one at it, i should add |
12:41.18 | alex_ | i sent a few version bumps to the list a few days ago btw.. are you guys not interested in that stuff or does it just take time to test and commit? |
12:43.23 | *** join/#uclibc sjhill (n=sjhill@real.realitydiluted.com) |
12:53.22 | blindvt` | alex_, i'm swamped with other stuff, sorry. |
12:53.50 | blindvt` | alex_, better put them into bugs.uclibc.org so i have a chance to find them when i return |
12:56.39 | blindvt` | alex_, what bumps are you referring to, exactly? |
13:01.07 | CIA-10 | 03aldot * r17905 10buildroot/package/dhcp/dhcp.mk: |
13:01.07 | CIA-10 | - provide -source targets for the 3 parts of the split dhcp. |
13:01.07 | CIA-10 | <PROTECTED> |
13:05.08 | alex_ | blindvt`: bash and genext2fs |
13:05.16 | alex_ | genext2fs was bumped i think |
13:11.05 | CIA-10 | 03aldot * r17906 10buildroot/package/bash/ (27 files): - bump to bash 3.2. Thanks to Alexander Rigbo |
13:11.39 | blindvt` | alex_, i'm using genext2fs myself, so rather want to try it out first. Please put it into the bug-tracker so i don't forget about it. TIA |
13:13.30 | alex_ | allright.. |
13:15.38 | alex_ | so.. you're aldot? |
13:15.57 | blindvt` | yes. And you're xride, IIRC? |
13:16.13 | blindvt` | ah, no. sorry |
13:19.56 | alex_ | genext2fs bump patch in mantis |
13:20.27 | alex_ | 1225 |
13:21.32 | alex_ | how nice that the patch became about 300kb just because it removes a big patch :| |
13:27.08 | alex_ | mmm, friday afternoon. |
14:15.52 | *** join/#uclibc elliotf__ (n=elliotf@adsl-71-140-174-98.dsl.pltn13.pacbell.net) |
14:16.35 | *** join/#uclibc ambroseL (n=bjb@router.fidus.ca) |
14:30.43 | *** join/#uclibc wrobbie (n=rob@cm74.kappa84.maxonline.com.sg) |
14:32.24 | CIA-10 | 03aldot * r17907 10buildroot/ (11 files in 8 dirs): - add skeleton to use busybox and mdev for a rather small skeleton |
15:03.11 | *** join/#uclibc landley (n=landley@c-71-199-114-26.hsd1.pa.comcast.net) |
15:04.31 | landley | Ok, I built an arm version of ulibc, and now I'm trying to build ldd and readelf and such to use in a cross-compiler. |
15:04.44 | landley | And it wants to link against an x86 version of uClibc, which I didn't build. |
15:04.53 | landley | Any way to tell it to link against the host libc? |
15:04.58 | landley | Other than hacking the makefile? |
15:12.17 | *** join/#uclibc blindvt__ (n=bf@M994P004.adsl.highway.telekom.at) |
15:19.51 | CIA-10 | 03aldot * r17908 10buildroot/package/busybox/busybox.mdev-just-read-9-bytes.patch: - just reading the first 9 bytes for the maj:min ought to be enough (TM) |
15:21.15 | landley | aldot: so we went to larger major/minors, and you want to make sure that a system that actually has 'em won't work? |
15:21.53 | landley | If you're going to hardcode a limit, make it 16 please. |
15:22.13 | blindvt` | exactly. |
15:22.42 | blindvt` | landley, i don't use LFS, just -boot c and have one disk, fwiw |
15:22.56 | landley | I missed... What? |
15:23.03 | blindvt` | nothing |
15:23.05 | landley | (I feel like I"m coming in halfway through a conversation...) |
15:23.20 | landley | blindvt: do you have enough memory on your system to run qemu? |
15:23.24 | landley | It's a marvelous thing, qemu. :) |
15:23.41 | landley | (Actually I've booted systems under qemu with -m 32...) |
15:23.46 | blindvt` | landley, yes, i have. -m 6 helps ;) |
15:24.26 | landley | LFS = large file support or Linux From Scratch? |
15:24.33 | landley | My first guess was #2. |
15:24.43 | landley | I now see -boot c as a qemu argument, before I was thinking "bootloader?" |
15:25.10 | blindvt` | largefile |
15:25.15 | blindvt` | :) |
15:29.15 | landley | Ah. |
15:29.37 | landley | I'm programming with 64 bit systems in mind, so I just assume large file. |
15:30.02 | landley | (Although in some places I'm lazy and go "look, a long will be 64 bits anywhere people actually care about feeding in a number larger than 4 billion". |
15:30.16 | landley | Which I realize isn't _quite_ the case, but in things like command line index parsing... |
15:31.17 | blindvt` | landley, btw.. may i ask you to glance at this inittab and tell me why busybox's init doesn't seem to get a working dev entry to proper do work? http://busybox.net/cgi-bin/viewcvs.cgi/trunk/buildroot/target/generic/target_busybox_skeleton/etc/inittab |
15:31.30 | landley | The way toybox's command line option parsing works is there's a global structure containing fields in a known order, so there's none of that varargs stuff. |
15:31.54 | blindvt` | landley, mdev is called and creates the nodes, but init doesn't seem to work nonetheless :/ |
15:32.04 | landley | because the nodes are created too late. |
15:32.07 | landley | Old problem. |
15:32.23 | landley | We open the file handles while parsing the inittab, so we can save the filehandle instead of the string. |
15:32.33 | landley | If you create the nodes _after_ parsing the inittab, it's too late. |
15:32.40 | blindvt` | landley, the initial /dev/ contains console, not using mdev (i.e. not remounting /dev/ to ram) works just fine |
15:33.05 | landley | Hmmm... |
15:33.29 | blindvt` | landley, resp. just booting with /dev/console and letting mdev populate the rest work, but not mounting a ramfs/tmpfs over /dev while init is already running |
15:34.08 | landley | blindvt: clarify, if your initial /dev contains /dev/console, and you don't mount anything on /dev, it works? |
15:34.11 | landley | (Even if you use mdev?) |
15:34.18 | landley | But if you mount a tmpfs on /dev, it stops working? |
15:34.19 | blindvt` | puzzles me a bit |
15:34.26 | blindvt` | exactly. |
15:34.31 | landley | Sounds weird, overmounting shouldn't screw anything up with existing filehandles. |
15:34.35 | blindvt` | let me try one again: |
15:34.48 | landley | It would have to be trying to open something new in the window between the mount and mdev running. |
15:35.05 | blindvt` | initially i have /dev/console, then, via inittab run mdev -s after / is remounted rw. works |
15:35.21 | landley | (My mdev didn't try to open anything out of /dev before running, but it's been modified a bit since I last looked at it...) |
15:35.35 | landley | The snapshot I have to clean up and apply to toybox is from the 1.2 timeframe... |
15:35.42 | blindvt` | initially i have /dev/console, then, via inittab mount tmpfs on /dev, run mdev -s . fails |
15:36.23 | landley | blindvt: because you told it to open /dev/null to associate a thingy with the new device, except it should have that filehandle cached? |
15:36.32 | landley | (Unless they fixed that to store the string since I last looked at it.) |
15:36.51 | *** join/#uclibc JockeHome (n=JockeHom@84-217-83-96.tn.glocalnet.net) |
15:36.53 | blindvt` | landley, i already looked if any change could have broken it, but didn't see anything significant that was changed since your version |
15:37.07 | landley | blindvt: by the way, trying to do a shell script within inittab is ugly and evil. |
15:37.12 | landley | Just do a shell script already. |
15:37.26 | landley | The order the lines are executed in is not guaranteed, and has already reversed once (in the 1.1 timeframe, I believe.) |
15:37.43 | landley | I suspect mdev is working fine. |
15:37.52 | landley | Afterwards, is /dev populated? |
15:38.01 | landley | By the way, what's the "failed" symptom here? |
15:38.06 | blindvt` | yea, i think it works fine but somehow init get's it wrong |
15:38.35 | blindvt` | "failed" in that rcS is run but i can't access the getty |
15:38.37 | landley | init is sort of evil... |
15:39.32 | landley | blindvt: I thought sysinit lines were run in _reverse_ order in our current init. Did that get switched again? |
15:39.42 | landley | (It's creating a linked list the "easy" way.) |
15:39.53 | landley | And there is NO SPEC for this sort of thing. |
15:40.02 | landley | I've looked. |
15:40.43 | landley | Could you confirm that the mount is run before mdev, and mdev run before your rcS? |
15:40.43 | landley | And why isn't the mdev thing _in_ your rcS? |
15:42.09 | landley | Finishing up the earlier thought: I already taught toybox to parse integers when I tell it an option is type # instead of type :, and now I'm teaching it about k/m/g/t suffixes. |
15:42.37 | landley | I figure that "head -n 4k" being understood isn't actually a _downside_... |
15:42.58 | landley | And it saves me a lot of effort for things like dd. |
15:46.26 | blindvt` | ah, the nice k/m/g/t stuff. like: |
15:46.28 | blindvt` | <PROTECTED> |
15:46.28 | blindvt` | <PROTECTED> |
15:46.28 | blindvt` | <PROTECTED> |
15:46.28 | blindvt` | <PROTECTED> |
15:46.29 | blindvt` | <PROTECTED> |
15:47.16 | blindvt` | where this refers to "EC 60027-2, Second edition, 2000-11" for your amusement ;) |
15:49.48 | blindvt` | where that metric_t is a dirty |
15:49.50 | blindvt` | <PROTECTED> |
15:49.50 | blindvt` | <PROTECTED> |
15:49.51 | blindvt` | <PROTECTED> |
15:49.51 | blindvt` | <PROTECTED> |
15:49.51 | blindvt` | <PROTECTED> |
15:49.51 | blindvt` | <PROTECTED> |
15:49.53 | blindvt` | <PROTECTED> |
15:49.55 | blindvt` | <PROTECTED> |
15:49.57 | blindvt` | <PROTECTED> |
15:49.59 | blindvt` | <PROTECTED> |
15:50.01 | blindvt` | <PROTECTED> |
15:50.03 | blindvt` | <PROTECTED> |
15:50.47 | landley | blindvt: I will never support kibibytes. Ever. |
15:50.50 | landley | It's a deeply stupid word. |
15:50.54 | landley | And I'm not going there. |
15:51.30 | landley | A kilobyte is 1<<10 |
15:51.33 | landley | megabyte 2<<20 |
15:51.41 | landley | er, 1<<20 |
15:51.46 | landley | gigabyte 1<<30 |
15:52.09 | landley | And terabyte can't be represented in "unsigned int" even on a 64 bit platform, due to LP64. |
15:52.20 | landley | Of course I'm using longs for option parsing. :) |
15:52.34 | landley | (Guaranteed to be the same size as a pointer, so my option parsing an work on type independent slots in the structure...) |
15:53.42 | blindvt` | i knew you'd hate it :P |
15:53.47 | landley | Looks like exabytes is all I'll ever have to care about, option-parsing-wise. |
15:53.53 | landley | I'm on record as previously hating it. |
15:54.07 | landley | (This is one of the long dragged out arguments I had with Eric Raymond, who bowed to the standard.) |
15:54.12 | landley | Despite the name being STUPID. |
15:54.29 | landley | Anyway, exa is 1<<60 and that's the biggest uint64_t can store. |
15:55.14 | landley | so roughly 1<<((index("kmgtpe",letter)+1)*10) |
15:55.53 | landley | so roughly 1<<((index("kmgtpe",tolower(letter))+1)*10) |
15:56.03 | blindvt` | except that there is no "index()" in my world, but ok |
15:57.09 | landley | c="kmgtpe"; x=index("kmgtpe",tolower(letter)); if (x) res=1<<((index("kmgtpe"-c,tolower(letter))+1)*10) |
15:57.11 | landley | ish |
16:01.37 | landley | Heh, ok, what does it mean to append a prefix to an octal or hex number? :) |
16:02.01 | landley | (Is understanding "head -n 070" a bad thing? :) |
16:02.19 | landley | head -n 0x7Fk has got to be just plain confusing, but really easy to program support for... |
16:02.43 | landley | Feed 0 to strtol's base field (autodetect, defauting to 10), then parse suffixes as normally... |
16:03.22 | blindvt` | populating a tmpfs /dev via rcS work fine, as does doing it manually from a shell. Looking at init.. |
16:03.44 | landley | blindvt: look into the order in which sysinit values are executed. |
16:03.59 | landley | It really was doing 'em in reverse order last I checked, which violates no spec I'm aware of. |
16:04.10 | landley | Order of these suckers within a class has never been guaranteed. |
16:04.15 | landley | Only ordering between classes. |
16:45.24 | *** join/#uclibc sjhill (n=sjhill@real.realitydiluted.com) |
17:17.09 | CIA-10 | 03aldot * r17909 10busybox/init/init.c: - whitespace cleanup and add a possible shrinkage suggestion |
17:41.31 | dalias | btw landley |
17:42.09 | dalias | 1<<((index(" kmgtpe",tolower(letter)))*10) might be a byte or two smaller than your code |
17:42.21 | dalias | if you can ensure that letter isn't space..which i guess you can't, but.. |
17:42.27 | dalias | 1<<((index("0kmgtpe",tolower(letter)))*10) should work |
17:48.57 | landley | dalias: if the compiler's smart enough to use inc for the +1 it should only be one byte, which you add to the string anyway. |
17:50.24 | dalias | what about 1024<<((index("kmgtpe",tolower(letter)))*10) |
17:50.38 | landley | dalias: that's an improvement, yes. |
17:51.18 | landley | And "my code" has wandered rather far away from my doodle above. |
17:51.35 | dalias | (if the compiler were smart it would use 2 bytes to load 1 and 5 bytes to load 1024 - xor,inc vs mov immed. but the compiler isn't smart :) |
17:52.38 | landley | dalias: this is not my problem. |
17:52.43 | dalias | *nod* |
17:52.57 | landley | All projects under FSF maintainership bloat to a stop and stall. |
17:53.06 | landley | If they're interesting, they then fork and somebody else throws out most of the crap. |
17:53.19 | landley | I can wait for gcc to fork away from the FSF again. Unlikely to take more than about 5 years. |
17:53.38 | landley | So predicting what future compilers may or may not do is something I try not to get too deep into. |
17:53.42 | dalias | well in the case of gcc i blame the cygnus ppl for bloating it to hell |
17:53.47 | landley | (Although "inc" has been a one byte instruction since the PDP-11 days.) |
17:54.00 | landley | dalias: I don't. |
17:54.09 | dalias | gcc 2.7.2.3 wasn't badly bloated |
17:54.16 | landley | dalias: I'm not going there right now. |
17:54.19 | landley | busy. |
17:54.20 | dalias | ok |
18:02.51 | CIA-10 | 03jocke * r17910 10uClibc/ldso/ldso/dl-elf.c: |
18:02.51 | CIA-10 | Fix dlopen for static apps. |
18:02.51 | CIA-10 | From Carmelo Amoroso. |
18:02.54 | *** join/#uclibc warly_ (n=warly@ASt-Lambert-151-1-14-151.w82-120.abo.wanadoo.fr) |
18:03.21 | *** join/#uclibc Ionic` (i=ionic-ms@ionic.de) |
18:04.03 | Ionic` | Hey, what to do if I'm getting "stage1/libgcc_s.so: undefined reference to `dl_iterate_phdr'" when bootstrapping GCC with uClibc 0.2.27? :) |
18:04.54 | Ionic` | I read something about adding "-D_GNU_SOURCE" to CFLAGS, will that do the job? |
18:11.49 | CIA-10 | 03aldot * r17911 10buildroot/package/l2tp/ (l2tp.mk l2tpd): - add runlevel script |
18:19.21 | landley | uClibc 0.2.27??? |
18:20.48 | landley | As for the dl_iterate_phdr thing, that was fixed ages ago. |
18:20.55 | landley | It's probably in 0.2.28. |
18:21.03 | landley | By which I assume you mean 0.9.28... |
18:22.47 | Ionic` | Eh sorry, I mean 0.9.27 ofc |
18:32.22 | CIA-10 | 03aldot * r17912 10buildroot/package/mtd/mtd.mk: - fix build for the host |
18:33.27 | blindvt` | 0.9.28.1, even |
18:44.01 | landley | Or the original proposed-by-maintainer ship date of 0.9.29 which was something like 9 months ago at this point... |
18:45.39 | Ionic` | Uhm well, is there any working solutions without recompiling uClibc? |
18:51.21 | landley | Ionic: not that I know of. You're trying to use a current buildroot with an old uClibc, I take it? |
18:52.24 | Ionic` | landley: hm in fact... no. I'm just trying to compile GCC-3.4.4 :) |
19:05.39 | JockeHome | Ionic`: Just add an empty dl_iterate_phdr to gcc. I don't think gcc will use the function |
19:07.07 | Ionic` | You mean somewhere in the libiberty-Source? |
19:07.18 | Ionic` | Or libgcc even |
19:07.56 | JockeHome | wherever it is needed, can't remeber where that is |
19:08.50 | Ionic` | Hm I'll look into some Gentoo-patches |
19:09.39 | JockeHome | or upgrade to .28 :) |
19:11.05 | Ionic` | JockeHome: I'd like to, but since the two versions are not binary compatible... |
19:11.05 | JockeHome | I see |
19:11.05 | JockeHome | well, you could add it to uClibc as well. |
19:11.17 | JockeHome | look in ldso/ldso/dl-elf.c |
19:11.46 | Ionic` | Mh it's just a little bit more complicated since I'd have to make my own new OpenWrt image for this and so on |
19:14.05 | JockeHome | Oh, well. It won't get any easier than that |
19:15.09 | Ionic` | Hm you know, the GCC i built in one Gentoo chroot is working and didn't through any error |
19:40.37 | JockeHome | Ionic`: because Gentoo has patches to fix this, great system :) |
19:51.30 | Ionic` | JockeHome: hehe... |
19:53.23 | Ionic` | JockeHome: though I cannot find any patches relating uClibc... weird |
19:54.21 | Ionic` | Maybe GCC |
19:55.46 | Ionic` | Neither, weird |
20:03.12 | *** join/#uclibc tchan1 (n=tchan@c-24-13-84-219.hsd1.il.comcast.net) |
20:16.26 | CIA-10 | 03jacmet * r17913 10buildroot/package/ (Config.in iperf/ iperf/Config.in iperf/iperf.mk): iperf package. Closes #0001224 |
21:08.56 | *** join/#uclibc ambroseL (n=bjb@router.fidus.ca) |
21:34.36 | CIA-10 | 03vapier * r17914 10uClibc/test/dlopen/dltest.c: cleanup a little |
21:35.24 | CIA-10 | 03vapier * r17915 10uClibc/test/dlopen/ (. Makefile dlstatic.c libstatic.c): make sure static variables are re-initialized everytime |
23:12.28 | *** join/#uclibc blindvt_ (n=bf@62.47.156.164) |
23:30.40 | *** join/#uclibc blindvt_ (n=bf@62.47.157.235) |