IRC log for #elinux on 20080105

00:05.17*** join/#elinux dijenerate (n=dijenera@72.51.111.182)
00:06.03*** join/#elinux likewise (n=leon@82-171-189-134.dsl.ip.tiscali.nl)
00:39.51*** join/#elinux dsaxena (n=dsaxena@static-64-65-140-161.dsl.pdx.eschelon.com)
00:51.58zx80userthe /etc/modules file appears not to work on busybox - is there an alternative?
04:20.43*** join/#elinux chouimat|busy (n=dieu@r2351064.cidc.net)
04:52.30*** join/#elinux chouimat|away (n=dieu@r2351064.cidc.net)
05:19.48*** join/#elinux chouimat|busy (n=dieu@r2351064.cidc.net)
05:53.48*** join/#elinux dsaxena (n=dsaxena@c-71-59-208-117.hsd1.or.comcast.net)
08:12.43*** join/#elinux fishing87 (n=charissa@CPE-124-178-129-157.wa.bigpond.net.au)
08:12.55*** part/#elinux fishing87 (n=charissa@CPE-124-178-129-157.wa.bigpond.net.au)
09:13.46*** join/#elinux AD-N770 (n=jep@24.Red-80-26-22.staticIP.rima-tde.net)
10:10.58*** join/#elinux CIA-1 (n=CIA@208.69.182.149)
10:32.48*** join/#elinux lyakh (n=lyakh@p57BD1BD5.dip0.t-ipconnect.de)
14:42.18*** join/#elinux zx80user (n=adrian@ip-81-1-89-80.cust.homechoice.net)
15:19.15*** join/#elinux GPSFan (n=kenm@12.10.255.246)
15:24.38*** join/#elinux fonz_ (n=tp@e178056096.adsl.alicedsl.de)
17:44.49*** join/#elinux Dr{Who} (n=mathewss@71-222-16-170.ptld.qwest.net)
17:48.02Dr{Who}having some problems booting on a via epia 5000 board. My setup works fine on qemu or an intel board. Kernel starts loading the init but it is as if the init process never starts. I have a simple "hello world" init binary staticly compiled to try and debug this. On the intel box or qemu it works fine i see it load the elf binary on my via board as its loading the binary it segfaults during the memory copy and i get a strange event sent to
17:48.02Dr{Who}the thread. I end up seeing a do_notify_resume with 0x04 or TIF_SIGPENDING.. then it just keeps doing this over and over
17:48.25Dr{Who}need ideas how to continue to debug this
17:49.15Dr{Who}2.6.22.10 kernel btw. i have patched in kdb to help debug
17:53.21Dr{Who}i also noticed that during the copy of the elf bin into user memry a clear_user is called on both boxes a page_fault is issued im presuming for dynamic memory allocation. But in the intel vs via the signal's re different. intel has a sighand_ctor where the via just goes direclty to error_code
18:21.06*** join/#elinux sstallion (n=sstallio@unaffiliated/icon)
18:21.08sstallionmorning guys
18:21.19sstallionmorning guys... do any of you know if a openssl compatible micro-lib exists out there?
18:22.43Dr{Who}my buildroot tree has a make setup for openssl
18:23.04Dr{Who}and i have no problems compiling it
18:23.42sstallionwell, i ended up not being able to use buildroot on this project :/
18:23.45sstallion(unfortunately)
18:24.07Dr{Who}well the make directory should give you any help you may need though
18:24.14Dr{Who}let me look at mine see if they do anything special
18:24.47sstallionthanks a bunch
18:25.21Dr{Who}ya a few patches in there
18:25.29Dr{Who}looks like they patch the configure
18:26.19Dr{Who}and the patch out building of the docs
18:26.36Dr{Who}maybe if you just used the patches and then cross compile should work
18:26.56*** join/#elinux __fonz__ (n=tp@e178033238.adsl.alicedsl.de)
18:28.11Dr{Who}looks like the main patch has lots of stuff for different processors and asmembly/compiler changes
18:33.09sstallionhrm ill give that a shot
18:33.12sstallionthanks again
18:34.54Dr{Who}also look at this http://landley.net/code/firmware/about.html
18:35.10Dr{Who}landley is on #uclibc he may have some ideas on openssl
18:35.43Dr{Who}i have not used his build system i use buildroot but it looks solid..
18:39.05sstallionhrmm i'll do that
18:39.40sstallionwe have a custom toolchain here so its a little tough to use buildroot or other scripts :/
18:40.03Dr{Who}ya.. well try that patch or at least parts of it
18:40.15Dr{Who}most look ok some parts are just changing path's etc
18:40.17sstallionwill do - i have a compile rolling on it now
18:40.19sstallionyeah
18:40.23Dr{Who}then build with that and see what breaks : c 0
18:40.43sstallionhrmm built clean
18:40.50Dr{Who}yay.
18:40.55Dr{Who}: c ) now does it work..........
18:41.01sstalliontests blew out... dammnit
18:41.24sstallionthe ironic thing is that libssl is only a compile time req' its not actually used on the target
18:41.28Dr{Who}as in openssl tests? or some of your own tests?
18:41.33sstallionopenssl tests
18:41.37Dr{Who}ahh.
18:41.51Dr{Who}well i dont know.. that may be tough. the openssl tests may also need fixing.
18:41.57sstallion*nod*
18:42.35Dr{Who}does your configure have an option to remove that dependancy?
18:42.44sstallionlooks like it
18:43.04Dr{Who}ya like --nossl or something
18:43.33sstallionoh, you mean on the dependant software? no
18:43.46sstallionits a hard dep
18:43.47Dr{Who}ah.
18:43.52sstallion(bastards)
18:44.10Dr{Who}so this software at what point does it break?
18:44.15Dr{Who}during configure? or make?
18:44.26sstallionits fully make driven, so during make
18:44.40Dr{Who}and the error?
18:44.57sstallionbasically that it depends on openssl (ie: cant find headers, cant link to libssl etc.)
18:44.59Dr{Who}i would say you can just modify the make to remove the -l for ssl
18:45.26Dr{Who}well #includes you would have to remove manualy if it truely does not need them
18:45.28sstalliontrue... i could just remove ssl support
18:45.34Dr{Who}and the libssl you can remove just modify the link options.
18:45.36sstallioneasy enough to add a NO_SSL macro
18:45.46Dr{Who}yep.
18:46.15sstallionhrmm maybe 50 source files... that may end up being easier
18:46.27sstallionseems a shame to trash a nice lean busybox build with openssl ;)
18:46.34Dr{Who}heh
18:46.47Dr{Who}so back to the openssl build itself
18:46.50Dr{Who}it did compile right?
18:47.10sstallionit did, but not using a true cross-compile
18:47.15Dr{Who}ahh
18:47.24Dr{Who}ya need to be cross compied
18:47.26sstallionthe target is a i386 (im on an x64) - so it doesnt really count ;)
18:47.28Dr{Who}or it wont like correctly.
18:47.31sstallion*nod*
18:47.38Dr{Who}its uclibc right?
18:47.48sstallionyup
18:47.52sstallioni386-linux-uclibc
18:48.10Dr{Who}ya so you have to build it against uclibc also then it will link correctly with your project
18:48.16sstallion*nod*
18:48.23sstallioni might look into disabling ssl altogether
18:48.27sstallionits not needed for this prototype
18:48.42sstallioni think it may just be an addon to the protocol were working with
18:49.00*** join/#elinux thraxisp (n=thraxisp@24.139.16.154)
18:49.10Dr{Who}ya probably. an option to talk ssl or clear text via tcp is common.
18:49.22sstallionahh interesting... its wanting to use lcrypt
18:49.38sstallionthis might be easier to replace just by providing an alternate libcrypt and swapping out some headers
18:50.16Dr{Who}well libcrypt is for basic encryption math md5's etc etc
18:50.19sstalliondoes uclibc provide a libcrpt?
18:50.21sstallionright
18:50.28Dr{Who}let me see
18:50.32sstalliontheyre relying on md5 and sha1 sums
18:50.41sstallion(just using openssl's version of it)
18:50.46Dr{Who}ahh
18:50.55Dr{Who}ok well that is easier to remove the dependancy then
18:51.01sstallionyup
18:52.02Dr{Who}libgcrypt
18:52.15Dr{Who}its an option on uclibc
18:52.32sstallionlooking in there now
18:53.23sstallionany idea of what the option is named?
18:54.05Dr{Who}sorry i ment that is for buidroot my bad
18:54.11sstallionah
18:54.13Dr{Who}uclibc is just libc
18:54.21Dr{Who}i dont think they have anything else.
18:54.28sstallionwell they add in a couple of extras... pthreads etc.
18:54.37Dr{Who}ok true that.
18:54.38sstallionalthough i guess technically you could call that part of libc
18:55.10sstallionhrmm
18:55.18Dr{Who}you know its not hard to make a .mk pacakge of your own for buildroot
18:55.52sstallioni think the part that dissapointed on buildroot for this project was the number of artifacts the owner considered unneccessary
18:56.04Dr{Who}then you can just grab the libs and bins you need and deploy it as you wish.
18:58.18Dr{Who}for example if i want to build something for my mips based tivo box i would use buildroot just to get a cross compiler and the libs i want then i can compile static if needed to and wala i have a bin for my tivo box.
18:58.35sstalliontrue
18:59.01sstallionhrmm looks like only one bit of source uses openssl for md5 and sha1
18:59.11Dr{Who}for that matter you can add --static and build your bin directly to x86
18:59.14sstallioni think it will be better just to find an alternate version of lcrypt
18:59.23sstallion*nod*
18:59.27Dr{Who}directly from any buildtree
18:59.48Dr{Who}it will be big but as you said its a prototype : c )
19:00.19Dr{Who}the only bugger wil be if your app dynamicly loads lib's i have never gotten that to work..
19:00.30Dr{Who}like if it has modules.. it loads.
19:00.36sstallionahh
19:00.40sstallioni dont think so... not with this
19:00.55Dr{Who}well then just add --static to your make and it should run on your box
19:00.57sstallioni just spent a week squashing all static bin's though ;)
19:00.59Dr{Who}regardless of lib's
19:01.15Dr{Who}heh
19:01.27Dr{Who}2 steps forward 1 back.
19:01.33Dr{Who}tis the name of the game
19:01.42sstallionamen to that
19:02.04sstallionstill, adding in openssl for f'n md5 is terrible - i cant believe these guys did that
19:03.16sstallioninteresting
19:03.23Dr{Who}so maybe just switch out the functions with libcrypt
19:03.23sstallionuclibc *does* have libcrypt
19:03.28Dr{Who}realy?
19:03.32sstallionyup
19:03.36sstallionhttp://www.uclibc.org/cgi-bin/viewcvs.cgi/trunk/uClibc/libcrypt/
19:04.08sstallioni'll be damned
19:04.13sstallionits built
19:04.19sstallionso this should be a quick fix
19:04.26Dr{Who}wow there ya go.
19:05.12sstallionjust need to convert to crypt.h
19:06.29sstallioncrap
19:07.17sstalliontheyre not using the crypt functs... theyre using openssl specific ones
19:07.51Dr{Who}well i would think they are simmilar so patching in calls to libcrypt vs openssl is probably not a huge diff
19:08.02sstalliontheyre pretty close
19:08.04Dr{Who}openssl lib is easy to use used it many times.
19:08.52Dr{Who}just dont need to init the lib and load your modules you would just call directly the crypt functions
19:09.36sstallioni might have to convert all of this into a crypt() call instead of the md5 primitives
19:45.59*** join/#elinux dsaxena (n=dsaxena@68.178.72.172)
19:52.34*** part/#elinux sstallion (n=sstallio@unaffiliated/icon)
21:16.19*** join/#elinux likewise (n=leon@82-171-189-134.dsl.ip.tiscali.nl)
21:48.51zx80userhas anyone here had any problems with the alsalib x-build on buildroot? The people on the alsa-dev list suggest my problems may be a broken set of paths and I wonder if anybody else has had this?
21:53.17*** join/#elinux ashes (n=ashes@2001:5c0:8fff:ffff:0:0:0:b3)
22:32.39*** join/#elinux _ccjoe_ (n=ccjoe@75-169-16-90.slkc.qwest.net)

Generated by irclog2html.pl Modified by Tim Riker to work with infobot.