02:08.15 | ambroseL | Hi, I'm compiling uclibc so I can statically link a small app to it. The uclibc build stopped with the following error: (hand-typed, sorry) In file included from timer_create.c:12: |
02:08.17 | ambroseL | kernel-posix-timers.h:29: error: parse error before "pthread_attr_t" |
02:08.19 | ambroseL | and there's more |
02:09.05 | ambroseL | I wonder if I need to include more features? I don't really want to include pthreads... I don't need pthreads to be able to use fork, do I |
02:09.06 | ambroseL | ? |
02:09.50 | *** join/#uclibc edrx (~user@201008007109.user.veloxzone.com.br) |
02:10.09 | ashes | im not sure. maybe experiment with the config a bit more |
02:21.52 | ambroseL | Well, ok. |
02:33.33 | ambroseL | I suppose pasting my .config here would be a big no-no, eh? |
02:33.42 | *** join/#uclibc simmo (~simmo@adsl-18-27.swiftdsl.com.au) |
02:41.14 | mjn3 | apparently someone added librt recently and didn't make sure things still built with threading disabled |
02:46.43 | ambroseL | That looks like the problem I saw |
02:48.18 | ambroseL | I was just starting to recompile uclibc with as few changes to the defaults as I could do to see if the problem was still there, but it is my background task so it's going a bit slowly. |
02:48.34 | mjn3 | ah... wait. no. looks like pthread.h wasn't previously included and was fixed later |
02:49.01 | mjn3 | http://uclibc.org/cgi-bin/viewcvs.cgi/trunk/uClibc/librt/kernel-posix-timers.h?rev=9722&view=log |
02:51.07 | ambroseL | I downloaded the 0.9.27 tarball, but the change you pointed out isn't in it. |
02:51.22 | ambroseL | Is there a later tarball? Or do I get to try svn? |
02:52.49 | mjn3 | you can either try svn or simply apply that patch to 0.9.27 |
02:53.00 | ambroseL | heh, true. Thanks mjn3 |
02:53.08 | mjn3 | all they did was include pthread.h |
02:57.35 | ambroseL | yay, it compiled! |
03:04.49 | ambroseL | Ok, it seems to compile my little application I need a special gcc that works with uclibc... that will have to wait for tomorrow. |
03:05.10 | ambroseL | Actually, if I'm statically linking with --nostdlib, do I still need the special gcc? |
03:07.49 | mjn3 | probably not. but make sure you build with the correct headers |
03:08.55 | mjn3 | certainly we used to build things all the time with the old gcc wrapper. there shouldn't be any problems for static |
03:09.20 | mjn3 | umm... depending on what it needs from libgcc of course |
03:13.03 | ambroseL | My app sits around waiting for network connections, and responds in ascii form over the network. It is designed not to touch the disk because we want to see if the nand flash is broken when a certain bug occurs. |
03:13.46 | ambroseL | So it is a simple app, no fancy stuff, basic open, read, write, close, opendir stuff to read from tmpfs and /proc |
03:14.41 | ambroseL | All in one file |
03:14.57 | ambroseL | printf |
03:15.38 | ambroseL | I'm going to see if I can stuff in a fork with alarm to execute risky tests like reading/writing nand |
03:15.57 | ambroseL | pipes to talk to child |
03:16.00 | mjn3 | i don't forsee any problems |
03:16.13 | ambroseL | Thanks very much mjn3, your advice is much appreciated |
03:16.24 | mjn3 | yw |
03:22.07 | mjn3 | grr... sjhill... |
03:30.41 | SpanKY | heh |
03:30.45 | SpanKY | delete the conf patch on you |
03:30.56 | SpanKY | at least i'm not the only one now ! :D |
03:31.04 | mjn3 | yeah... |
03:31.05 | mjn3 | heh |
03:31.39 | mjn3 | he's focused on what he's doing and not seeing the bigger picture i think |
03:32.05 | mjn3 | can't complain too much. he did most/all of the initial mips port way back when |
03:35.20 | SpanKY | you ever get a min to rip me a new one for the coreutils/stat.c ? |
03:36.29 | mjn3 | afraid not. wound up doing some toolchain stuff (as you know) on saturday and spent sunday with carolyn since she's working all of this coming weekend |
03:37.09 | mjn3 | actually, i'm starting to work on my libm rewrite again so the stat applet has dropped a bit on the priority list |
03:38.53 | SpanKY | ah yeah, libm is huge |
03:38.57 | mjn3 | that and i've _been_ meaning to tweak things so that the prebuilt locale stuff was arch-independent (barring endianness of course). then i'd write a util to convert to the other endian. shouldn't really take lonk |
03:38.58 | SpanKY | still looking forward to that :) |
03:39.37 | mjn3 | and psm was asking me about an iconv implementation for the stub locale build |
04:03.12 | *** join/#uclibc andersee (~andersen@12.158.229.94) |
05:06.16 | *** join/#uclibc n2n (nobody@202.63.233.18) |
05:49.48 | *** join/#uclibc ambrose (~bjb@ns1.linuxbutler.ca) |
06:43.59 | *** join/#uclibc courtc (~courtc@adsl-217-10-197.asm.bellsouth.net) |
09:00.06 | *** join/#uclibc woglinde (woglinde@pD9E7FD8A.dip.t-dialin.net) |
11:31.53 | *** join/#uclibc xilun (~xilun@xilun.rox0r.net) |
13:13.06 | thomasez | Manuel? |
13:19.08 | mjn3 | you rang? |
13:19.39 | mjn3 | thomasez: make it quick. i'm leaving for work in < 20 minutes |
13:21.42 | thomasez | the lzma you have used and the sdk are not really compatible. |
13:21.49 | thomasez | and mine are neither. |
13:22.20 | thomasez | but it's in line with other embedded usage of lzma. |
13:22.23 | thomasez | happy trip. |
13:22.34 | thomasez | the doorbell rang. |
13:24.21 | mjn3 | not compatible in what way? storage format? api (such as it is) ? |
13:56.41 | thomasez | you can't uncompress if it's compressed by another library. |
13:57.05 | thomasez | there are three, the stream/zlib I use, the lzmalib, which you use and the lzma SDK. |
14:16.23 | mjn3-work | ok. what are the differences between the 3? i know that the stuff i looked at had no real error checking for instance |
14:19.49 | mjn3-work | when i originally q |
14:20.19 | mjn3-work | when i originally looked at it, the lzmalib stuff closely matched the sdk |
14:34.35 | thomasez | I just compiled the lzmalib compressor and lzma sdk stuff and tried compressing/decompression teh same files and it just didn't work. |
14:35.38 | thomasez | and we went the zlib/stream route because it was the easiest. |
14:39.07 | *** join/#uclibc Sgt-Donan (~Loutre@lns-vlq-17f-81-56-167-49.adsl.proxad.net) |
14:40.47 | thomasez | and it works. |
15:50.16 | *** join/#uclibc ambroseL (~bjb@67.71.251.210) |
16:44.39 | *** join/#uclibc andersee (~andersen@es-gateway.directv.com) |
16:44.48 | mjn3-work | andersee: morning |
16:48.36 | andersee | mjn3-work: morning |
17:03.27 | ambroseL | Neat! Going forward from my work last night, I compiled my little program against uclibc, and had unresolved symbols __u{div,mod}si3 |
17:03.34 | ambroseL | I googled, and saw those are in libgcc |
17:03.49 | ambroseL | Tried linking against the libgcc in our existing toolchain |
17:04.01 | ambroseL | and the program compiled, linked, and even runs on the target! |
17:04.30 | ambroseL | fun, fun, fun! |
17:05.21 | andersee | ambrose: congrats |
17:05.32 | mjn3-work | glad to hear it. grabbing the math support funcs from libgcc should be fine. there are other things in libgcc you need to avoid though |
17:05.48 | mjn3-work | which is one of the reasons we build uClibc toolchains |
17:06.31 | ambroseL | yes... I will try that eventually |
17:06.57 | mjn3-work | guess i should do a klibc comparision again as well |
17:07.36 | mjn3-work | why they ever started klibc i'll never know |
17:07.43 | ambroseL | So my app in it's new form (normal gcc/uclubc) is just over 200KB vs over 500 KB (both stripped) compiled with full gcc/glibc |
17:07.47 | ambroseL | static link |
17:08.17 | ambroseL | Although, the md5sum stuff was in the non-uclibc version and has been removed from the uclibc version |
17:08.40 | mjn3-work | there's md5sum code in busybox |
17:08.56 | ambroseL | Hmmm :-) |
17:10.15 | ambroseL | And we have included it in our build... but my app needs to use a library. |
17:10.35 | ambroseL | my app must avoid going to disk for anything - shared libs, other executables, etc |
17:11.09 | ambroseL | I've been told by our senior guy to just leave it out for now. Oh well. |
17:37.48 | Lethal | mjn3-work, klibc was initially intended for moving out things like ipconfig and the do_mounts stuff for initramfs. so the reasons for starting it are quite sensible, but I dont see a lot of point in building on it to any huge extent as a general system thing. |
17:49.24 | mjn3-work | Lethal: you're missing the point. you can build uClibc in a stripped-down mode that is approximately same size (static), less buggy, and more featureful |
17:50.32 | Lethal | mjn3-work, and did you point this out to hpa? |
17:51.39 | mjn3-work | i never did. both andersee posed the question of "why start another small libc project" on the lkml at the time. so did fefe of dietlibc |
17:52.49 | Lethal | ah, don't remember that thread. suppose i'll grep later. |
17:54.37 | *** join/#uclibc ashes_ (~ashes@modemcable202.48-130-66.mc.videotron.ca) |
17:58.28 | *** join/#uclibc CIA-10 (~CIA@flapjack.navi.cx) |
18:32.37 | vapier | someone feel like lending me their GNU as ninja skills ? |
18:33.02 | vapier | i got this package which defines setjmp.S ... inside it defines both 'setjmp' and '__sigsetjmp' |
18:33.22 | vapier | binutils 2.15.92.0.2 assembles it just fine but 2.16 errors out |
18:34.28 | vapier | http://rafb.net/paste/results/wYRkO549.html |
18:34.49 | vapier | 2.16 (understandbly) aborts because of the .endp without a .proc for __sigsetjmp |
18:35.06 | vapier | if i put '.proc __sigsetjmp' just before the __sigsetjmp: label, it assembles fine under both |
18:35.25 | vapier | my question is, it shouldnt affect the final binary right ? |
18:48.14 | *** join/#uclibc GyrosGeier (~geier@port-212-202-73-190.dynamic.qsc.de) |
19:04.07 | ambroseL | Eek I'm out by a decimal place again. The program is actually a bit over 20KB stripped, down from 500KB stripped. :-) |
19:09.18 | mjn3-work | heh |
20:28.14 | *** join/#uclibc ashes (~ashes@modemcable202.48-130-66.mc.videotron.ca) |