<Textmode>
scratches head, and wonders where he put the xburst toolchain.
<bartbes>
if you built it you should be able to find it
<bartbes>
it's huge
<Textmode>
yes, yes it it.
<Textmode>
ah, it was on a seperate partition, which wasn't mounted.
<Textmode>
(and yes, it it the only thing on that partition.)
<bartbes>
and that partition is now full
<bartbes>
and it's multiple TBs
<Textmode>
bartbes: practically, 130mb free. :P
<tuxbrain>
urandom still arround in this channel?
<xiangfu>
wolfspraul: Hi
<wolfspraul>
ah hi.
<wolfspraul>
I setup a separate buildhost machine, fidelio.qi-hardware.com
<wolfspraul>
if you send me your public key I give you root access and create a user account too
<xiangfu>
wolfspraul: great.
<wolfspraul>
it's a very cheap server, only 29 EUR / month (on sale :-)) so it's not very powerful but I think still well suited as a build server. 64-bit athlon, 300gb raid-1 disk, etc.
<wolfspraul>
anyway before we upgrade hardware, we should first understand our real needs, so I thought I start with this
<wolfspraul>
I can also use it as a safe sandbox for things like the schematics diffs generation
<wolfspraul>
anyway it's there now. I am running a make with config.full_system right now (in my user account)
<wolfspraul>
of course anybody else who wants to build qi related stuff is welcome to use it too
<xiangfu>
ok.
<wolfspraul>
xiangfu: so do you want an account?
<wolfspraul>
I may have your public key already somewhere, can copy from there...
<xiangfu>
wolfspraul: how much time build the openwrt in this server.
<xiangfu>
wolfspraul: I can send you again. it's ok.
<wolfspraul>
don't know yet [build time]
<wolfspraul>
we can also use it later to build the entire world (all packages), which we need to do because upstream won't turn off patented codecs in their builds
<Textmode>
I've forgotten how to use the xburst toolchain :/
<wolfspraul>
actually I am wondering how well those images work :-)
<wolfspraul>
we cannot use them because they include patented codecs
<wolfspraul>
but they are definitely being built, and it's built in the way openwrt handles the 'backfire release' (whatever way that is)
<wolfspraul>
so how are these images built? how can we only derive from these images, without having to worry about each individual package revision?
<wolfspraul>
maybe they don't have any packages from feeds at all? feeds are not included in the release system? the rootfs tarball is only 1.5 MB...
<xiangfu>
the backfire only have thress packages repos:
<xiangfu>
we have six package repos.  see openwrt-xburst/feeds.conf
<xiangfu>
the .config can select or disclect which packge to build or not. but it can not select the package version.
<xiangfu>
so if user want build a backfire image in his own pc. he much checkout the packge revision to the date : Aug-27. (like the link you send to me"
<xiangfu>
not the last revision.
<wolfspraul>
it works with dates? you cannot pin some revision in the package feed repositories?
<xiangfu>
there are Aug-27 have five commit: @22823 ~ 22818. I don't know which one is for Aug-27 release. but definetly one of them.
<xiangfu>
there are five commits in Aug-27, in openwrt/packge svn.
<xiangfu>
there are five commits at Aug-23 in openwrt/package svn server. (sorry for my bad english :(
<wolfspraul>
where do those package versions come from?
<wolfspraul>
I'll just go through the compile process to understand it...
<wolfspraul>
xiangfu: what's the result with your latest image then? what works, what doesn't work?
<xiangfu>
wolfspraul: sorry. not test yet. will test it today.
<xiangfu>
wolfspraul: I cann't find out which revision of openwrt.org/backfire/10.03.1-rc3/xburst/packages/ use.
<xiangfu>
the package revision is output by "./scripts/feeds update" when he/she compile.
<wolfspraul>
that means our images are not reproducible
<wolfspraul>
because we discard the versions of the individual packages? well we are publishing a BUILD_LOG file so it's probably in there, but is there any effective way then to re-create our release images? maybe not...
<xiangfu>
wolfspraul: the BUILD_LOG is make outout. not include the ./scripts/feeds output.
<xiangfu>
wolfspraul: but we can check you current packge revison by .:
<xiangfu>
cd ./feeds/packages && svn info
<xiangfu>
we can.
<xiangfu>
for example: if you want reproduce my last build. you should:
<xiangfu>
cd ./feeds/packages && svn co -r 22968
<xiangfu>
not run "./scripts/feeds update -a" . because when you use "./scripts/feeds udpate -a" it will update feeds/package to r22976
<xiangfu>
I will try to write a script file for reproduce  openwrt release image.
<tuxbrain2>
bartbes: is the lua/love in the qi repo yet?
<qwebirc75230>
hi  some one in here?
<wolfspraul>
wow that was quick
<kristianpaul>
wolfspraul: i want compile openwrt stuff in your server
<wolfspraul>
do I have your public key somewhere?
<kristianpaul>
my cmputer is so slow righ not and still building since yday at night...
<kristianpaul>
wolfspraul: projects
<wolfspraul>
well I'm not sure the server is much faster :-)
<kristianpaul>
wolfspraul: indefero
<wolfspraul>
it's a single core athlon
<wolfspraul>
I am building config.full_system right now, started maybe 5h ago, still running
<kristianpaul>
gmm as my computer
<wolfspraul>
oh it's definitely good to have a server
<wolfspraul>
use screen, leave stuff running
<kristianpaul>
yesh !
<wolfspraul>
we need to ramp up the servers and builds anyway
<kristianpaul>
sure actually my dekstop coputer is powerfull than my web/blog/mail/other stuff server..
<kristianpaul>
but i cant do all by the comannd line (may i need reconsider that) when building openwrt..
<kristianpaul>
well lets wait your to see your performance timmings
<kristianpaul>
need new server a lab PSU ohh my god so much money !!!
<wpwrak>
gaah. way too much beer for a tuesday :-(
<wpwrak>
tuxbrain2: nice video !
<kristianpaul>
wolfspraul: that SDK you mentioned i hope will save time for this building, also i asked in openwrt-dev and ther is not such us auto build package thing
<wolfspraul>
kristianpaul: account name 'kristianpaul' ok?
<kristianpaul>
wolfspraul: well ok
<wolfspraul>
there is also something called 'image builder' - it's big but can save lots of time
<wolfspraul>
not good? what do you prefer?
<kristianpaul>
i dont sait it is not good :)
<kristianpaul>
i actually was hoping some auto packaging tool (i mean from source)
<kristianpaul>
but i know is ask too much
<kristianpaul>
and risk quality in packages
<kristianpaul>
wpwrak: morning !
<kristianpaul>
wpwrak: so mmap is dangueros? :)
<wpwrak>
kristianpaul: it's powerful. with great power comes great responsibility :-)
<tuxbrain2>
wpwrak: you passion for beer another reason to love you :), thanks  for the grettings master :)
<wpwrak>
tuxbrain2: still twiddling my thumbs with UPS/mail boxes. i copied you on the retransmission. (the first one, went only to them, without copy)
<wolfspraul>
kristianpaul: try it out kristianpaul@fidelio.qi-hardware.com
<wolfspraul>
also added you to root's authorized_keys
<wpwrak>
tuxbrain2: ah, wish the girls would love me for the beer :) usually, they just give my belly a critical glance
<wolfspraul>
my plan for this server is 'buildhost', let's say 'medium security' :-)
<wolfspraul>
so I see the whole thing as a big /tmp folder
<wolfspraul>
we can use it as a sandbox to run scripts like the schematics diff too, and similar stuff
<wolfspraul>
like a 'cpu + /tmp storage in the cloud'
<wolfspraul>
please don't store any valuable data on it, I am not doing any backups
<wpwrak>
let's see if they're already inefficient enough to bring the machine to its knees :)
<tuxbrain2>
wpwrak: I will see the mbe guy today I will comment, regarding the girs... well you know you have allways a bald, fat beared man waiting for you in spain :P
<wolfspraul>
also I will be more generous on that machine playing with testing, even unstable packages. for example I am on squeeze right now.
<wpwrak>
tuxbrain2: thanks. this will improve my ability to deal with rejection ;-))
<kristianpaul>
wpwrak: i'm in
<wpwrak>
wolfspraul: bte, the scripts should be good to use now in their new home. i did ben-wpan and xue without trouble. Makefile shows you how to use things, including some environment variables to make make the generated Web page nicer.
<kristianpaul>
wolfspraul: ^
<kristianpaul>
wpwrak: but is okay used it?, i just dont want end doing a kernel module that will drop me perfomance, what are those responsabillites btw?
<wpwrak>
kristianpaul: you just need to get the registers right. it's not more dangerous than writing anything that lives inside the kernel.
<kristianpaul>
wpwrak: sure i'll take care, actually i just want readdata most of the time, may be let a led for ocasional debugging
<wolfspraul>
it also describes the function of the server, lack of backup etc.
<kristianpaul>
wolfspraul: cool :)
<wpwrak>
kristianpaul: for users, it has the drawback that they need to run it as root and that there's no useful privilege separation (if you care about such things). on the technical side, you also don't have access to interrupts.
<wolfspraul>
feel free to use the machine aroudn the clock, if you have any useful load for it I'm glad to help
<wpwrak>
wolfspraul: you should have named the server "dirt" or such :)
<kristianpaul>
is not that defauklt (running root ) in openwrt wpwrak ?
<wolfspraul>
just please remember /tmp etc. - no backups, the way it's run downtime may be unavoidable sometimes
<kristianpaul>
ok
<wpwrak>
kristianpaul: dunno. i always run things as root. if you add up all the time it takes you to go back and put a sudo in front of whatever you're trying to do, i'm sure there's a whole year of your life you're wasting :)
<kristianpaul>
wpwrak: i'm not sure how usefull will be get 2millions interrupts every second..
<kristianpaul>
not for me :)
<wpwrak>
kristianpaul: 2 MHz interrupt frequency .. in your dreams :)
<kristianpaul>
wpwrak: i use sudo on other things i just notice opnewrt dotn encorage me to do it so
<kristianpaul>
sudo and non privileges users.
<wpwrak>
wolfspraul: now we just need a bit of simple web space, somehow connected to "projects" :)
<wolfspraul>
you mean for the schematics diffs?
<wolfspraul>
kristianpaul: I will also do ipv6 tests on this server. the dns already has a aaaa record for it, and a ipv6 reverse dns is equally setup
<wolfspraul>
so before going live with ipv6 stuff on other servers, I'll do experiments on this one
<wolfspraul>
a /64 ipv6 subnet is assigned to the server and packages should be routed to it, although I have never tried and don't know what would happen with such packages once they reach the kernel :-)
<wpwrak>
wolfspraul: manually uploading things into the wiki is just way too painful
<kristianpaul>
wpwrak: #Â Â AMD 64 3700+, 2 GB DDR400 RAM, 2 x 300 GB SATA (Software-RAID 1)
<kristianpaul>
wpwrak: sale? but you pay i moth by moth how is that?
<wpwrak>
wolfspraul: cpu looks decent. memory is a bit on the low side, though. lots of ram reall helps to speed up things.
<kristianpaul>
interested
<wolfspraul>
yeah but they have special etc. they need to fill their new mega datacenters...
<wolfspraul>
wpwrak: please explain me in more detail what you mean with 'web space attached to projects'
<kristianpaul>
wpwrak: so planning have loop filesystem in ram?
<wolfspraul>
what I planned to do was the schematics diff stuff, that is clear. the url needs to be relative to project names somehow, it all should work automatically for any kicad project
<wolfspraul>
we already have /people/ storage on downloads, just to dump stuff on a server. but you have that at almesberger.net too, no?
<wolfspraul>
so what more do you mean?
<wpwrak>
wolfspraul: oh, just a place to put the sort of stuff i currently keep on my server. checking it into git often doesn't give us the right kind of access from the web (see the feature request we had)
<wolfspraul>
kristianpaul: let's use this machine first, we can always upgrade, and computing power gets cheaper...
<wpwrak>
wolfspraul: ah, people sounds good. do i have that ?
<wolfspraul>
the raid1 will also slow it down, probably, for build style loads (too many writes)
<kristianpaul>
wolfspraul: sure i will use at much i can
<wolfspraul>
you prefer DAV or ssh/scp?
<wolfspraul>
if you want DAV, email me the password you want
<wpwrak>
(raid1) you can either make the storage system fast or you can add ram until you don' care about the disks. the latter tends to be cheaper for moderate loads :)
<wolfspraul>
otherwise I add your pubkey
<wpwrak>
definitely ssh/scp :)
<kristianpaul>
indeed
<wpwrak>
tuxbrain2: mboxes woke up. pickup should happen today.
<wolfspraul>
wpwrak: do you prefer werner or wpwrak for your people folder?
<kristianpaul>
wolfspraul: how i do crate a project in the qi indefero?
<kristianpaul>
i think is time start uploading some code :)
<wolfspraul>
let me give you super powers :-)
<kristianpaul>
wee :)
<wpwrak>
wolfspraul: "werner" would be good
<wpwrak>
kristianpaul: now you can write something with mmap there, too :)
<wolfspraul>
kristianpaul: I made you 'staff' for now, there is also admin but I need to lookup how to enable that.
<wolfspraul>
with staff rights you can definitely create projects
<wolfspraul>
I am kind of pushing out to add a 'create project' web form because I hope Indefero will get to it before I do :-)
<kristianpaul>
there is  recommende way to create project names, i noticed wpwrak used the prefix ben- wich is usefull to indentify ti is intende for the nanontoe
<wpwrak>
wolfspraul: works. thanks ! can i also delete files in some way ?
<wpwrak>
(except for overwriting)
<kristianpaul>
remote command?
<kristianpaul>
i guess
<kristianpaul>
ssh ww-data:@qi:werner rm -rf foo
<wolfspraul>
wpwrak: ssh should work. and just in case I added your pubkey to the root account as well.
<wpwrak>
kristianpaul: ah, that works. thanks :) getting to a shell was of course the first thing i tried, but when it didn't move, i thought it wasn't available. but now i see that it's just a little slow.
<kristianpaul>
:)
<wpwrak>
wolfspraul: you're making me worry about the security of my private key :)
<kristianpaul>
lol
<wpwrak>
actually, real per-user accounts would be better. less risk of mistakes.
<wpwrak>
wolfspraul: btw, did you find an answer to your question about breakout boards in my mail about the inrush current ?
<wolfspraul>
hmm :-
<wolfspraul>
:-)
<wolfspraul>
I was thinking...
<wolfspraul>
the answer seems to be at the end
<wolfspraul>
but I am not entirely sure
<wpwrak>
;-) there's basically not a lot you can do from the outside. except to be careful about not making things worse.
<wolfspraul>
in your experiments you had the LCM removed. will it help if it's on?
<wolfspraul>
what does this mean for ben-wpan? is it doable over sdio?
<wpwrak>
i think it should help a little. there's a lot of capacitors on there
<wolfspraul>
what does it mean for the breakout board? should we add some inductors there?
<wpwrak>
i think ben-wpan should be fine. i'll just "precharge" the caps. the ohmic resistance of the chip is very high when it comes out of reset, so it doesn't discharge the caps
<wolfspraul>
xiangfu: remember we wondered why your 09-06 rootfs was smaller than mirko's 08-26? I diffed it and it seems to be mostly Qt related. your image is missing many Qt things compared to 08-26.
<kristianpaul>
wpwrak: do you mind if i use ben-gps for the project i'm working on?
<wpwrak>
for the breakout board, it depend on how much the caps on the lcm help. if they are sufficient to reduce the drop to safe levels, then the only thing to worry about in breakout boards would be no to add capacitace - or put anything of that sort behind a bead
<wolfspraul>
if you want to I email the diff to you
<wpwrak>
(bead/inductor)
<wpwrak>
kristianpaul: not at all. i don't own the ben-* namespace anyway :)
<xiangfu>
wolfspraul: yes. I also found that.
<wpwrak>
(inrush) if the lcm isn't enough to prevent disaster with discharged caps, you also need to worry about the resistance between VDD and GND.
<wpwrak>
of course a voltage drop of ~ 1V is a lot. well beyond all specifications. so at least in theory the lcm won't be able to help. in practice, it might, though. also it may also depend on what the machine is doing at the time a board gets powered up.
<wpwrak>
(won't be able to help) because it can, best case, only about double the capacitance on the "good" side. so you would still get too deep a drop.
<xiangfu>
wolfspraul: I am try to debug the uClibc 0.9.32. first I start with objdump dgclock. the compare the 0.9.32 and 0.9.30 result.
<wolfspraul>
also try strace maybe?
<xiangfu>
yes.
<wpwrak>
grumbles in the general direction of those "reduced" tools and environments
<wpwrak>
well, uclibc is actually the least of my worries. busybox and dropbear are a different story, though.
<wolfspraul>
xiangfu: also I think in openwrt, you can turn on debug symbols globally. of course need to rebuild everything.
<wolfspraul>
maybe cut it down to just dgclock and dependencies first, and use the buildhost if it helps
<xiangfu>
hmm... ok. but first I will start with strace and objdump :)
<wolfspraul>
kristianpaul: building the config.full_system image took about 9h on the server
<wolfspraul>
I don't think the server is memory bound right now, maybe disk or CPU. hard to say. could try to turn off the raid1, but I think for now it's good enough. server works :-)
<wolfspraul>
kristianpaul: actually I have an easy idea. the /home partition should be ext2 not ext3. next time you have nothing running let me know then I can quickly downgrade that.
<kristianpaul>
wolfspraul: do it now
<kristianpaul>
i can wait for your is better now that theare few people
<kristianpaul>
i'm out , take your time and let me now
<wolfspraul>
now? he, OK. one sec.
<wolfspraul>
kristianpaul: maybe i can tech Linux software raid to favor performance over safety, instead of just turning it off
<wolfspraul>
will check that tomorrow, maybe a setting somewhere...
<wolfspraul>
we can still squeeze out some performance I'm sure
<kristianpaul>
sure no hury
<wolfspraul>
also I can try to upgrade the memory and mount /tmp in ram
<kristianpaul>
wolfspraul: linux raid (faik-raiid) is not good in perfonrmance indeed
<wolfspraul>
kristianpaul: I'm done /home is ext2 now
<wolfspraul>
makes more sense, should be a bit faster
<tuxbrain>
I'm in trip right now, but I solved the problem with greecee in the shop
<tuxbrain>
try again please and accept my apologizes
<kristoffer>
roger will do
<tuxbrain>
urandom__: I have read you were able to run Supertux! and asking for interest in the wiki forum sorry I miss that, yes I will love to have it in jlime and Openwrt repos :)
<tuxbrain>
ok guys must leave, anything until saturday better by email
<tuxbrain>
don't believe this tuxbrain2 :) is just a ghost
<tuxbrain>
c u (all)
<wpwrak>
tuxbrain2: package is on its way
<wpwrak>
grrr. i misunderstood how the transceiver's spi interface works. can't share miso and mosi after all :-( on the occasions where i saw it work, the signal shapes were quite horrible. ah well, back to the drawing board.
<emeb>
trying to tri-state them?
<wpwrak>
emeb: well, i thought that the chip would tristate MISO when it's not sending useful data. but i misread that paragraph and it only tristates it when nSS is inactive. so it's normal SPI after all.
<qi-bot>
[commit] Werner Almesberger: Found out that we can't share MISO and MOSI after all. Redesigned the atusd http://qi-hw.com/p/ben-wpan/3533e04
<GNUtoo|laptop>
hi
<GNUtoo|laptop>
I bought a GNU/Linux Magazine France today
<GNUtoo|laptop>
and there was an article on milkymist
<GNUtoo|laptop>
roughly they told that GNU stuff like gcc is kind of complicated to port
<GNUtoo|laptop>
to something like latice32
<GNUtoo|laptop>
I bet they plan to change that, for instance with the new license exception of GCC
<GNUtoo|laptop>
but I wonder how much time it will take
<wpwrak>
kristian1aul: btw, how did things go with the new acid ?
<kristian1aul>
wpwrak: dint try yet
<kristian1aul>
i need a thin PCB
<kristian1aul>
current will not fit on the slot
<kristian1aul>
wpwrak: your PCB is thin by default or you did some work before?
<kristianpaul>
i wonder why digikey dont ship in first class mail USPS and i wonder is asking for it will change something my bad feelins about a NOT Sr we DONT ...