<xiangfu>
modify the .config before the package build works fine.
<xiangfu>
kyak, only the URL is wrong. I will fix it.
<whitequark>
several months ago I've asked some stupid questions about the proper way of writing IPU drivers, and someone clever, whose nick I cannot remember now, answered here
<whitequark>
if that was you, then reply please :)
<kyak>
and there is no config.log.. was it the message from cmake?
<kyak>
dvdk: btw, perhaps it also makes sense to rewrite the Makefile, cause it was written in days when there was no cmake.mk
<kyak>
i don't know if this would help
<dvdk>
kyak: current cmake.mk won't help, doesn't support some tricks we're doing. would have to recode that. Â
<dvdk>
s/recode/add
<dvdk>
btw there's the log:
<dvdk>
CMakeFiles/CMakeError.log
<dvdk>
"libdl.c:(.text._dl_parse_relocation_information+0x1dc): undefined reference to `TLS_DTPREL_VALUE'"
<dvdk>
that's the problem.  new uclibc?  no TLS support?
<kyak>
new uclibc - for sure. and don't know what the hell TLS is
<kyak>
but it looks like the problem is on libdl side anywau..
<dvdk>
"thread local storage" (maybe?)
<dvdk>
not necessarily, maybe we need to link another lib when using libdl?
<dvdk>
but libdl is used everywhere.  if this is broken, you'd have many more build problems
<kyak>
maybe it's used, but "TLS" fucntions are not used
<dvdk>
no, the problem comes from libdl: it might be that libdl now uses libpthread and we'll have to tell plplot to link that, too during detection
<dvdk>
doing my next try, building plplot at least on master (enabled fortran support, maybe that's needed?
<dvdk>
)
<wpwrak>
DocScrutinizer: (EL PCB tester) hmm, how exactly would that work ? and where do you get such a display ?
<DocScrutinizer>
wpwrak: the (usually 400Hz though you can use higher freq) AC causes some flourescent substance to shine up by exciting it via capacitive AC field. I dunno where to get those "displays" maybe revamp backlights as used in jets and tanks, or sth
<wpwrak>
phew. now there's a piece that's hard to source :)
<DocScrutinizer>
I'm sure 3M or sb is selling this stuff by the meter nowadays
<DocScrutinizer>
maybe you even can get paint to apply to a glas covered with the gnd electrode SnO2
<DocScrutinizer>
or was it ZnO2? the transparent stuff used in LCD
<DocScrutinizer>
hard to source - maybe. Hard to build - don't think so
<DocScrutinizer>
if you got no better ideas and materials, then cover the whole PCB with a thin film of water with duh lakmus or phenolphtalein, and place a semi transparent conducting thing on top, then do electrolysis by feeding on pad under test with like 5V and see where other pads start "boiling"
<DocScrutinizer>
wpwrak: got the basic idea?
<DocScrutinizer>
it's not a QA tester but a mere RE tool
<DocScrutinizer>
for complex multilayer PCB to analyze the connections
<DocScrutinizer>
once you got the BOM, component placement, and this connection list, it's quite easy to feed this to $kicad and get a proper schematics
<DocScrutinizer>
soory for this being somewhat anti-topic here ;-D
<wpwrak>
DocScrutinizer: hmm. or maybe place the PCB in a vacuum, heat it, inject a high voltage, and watch electrons fly off it, CRT style :)
<wpwrak>
DocScrutinizer: i wonder how they actually do the electrical testing of PCBs in industry. bed of nails seems ... messy
<DocScrutinizer>
wpwrak: already thought about it, use front 'half' of a plasma screen
<DocScrutinizer>
wpwrak: they know their PCB, so have to pin-contact only very few known pads and probe for connect/isolation on a few pairs
<DocScrutinizer>
can be done with one pair of CNC probes
<wpwrak>
okay, that's for large volumes, where you make a custom tester. but for smaller/prototype volumes ?
<wpwrak>
ah, cnc probes. that would be slow then. but probably okay for small volumes
<wpwrak>
been trying to figure out a way to do such things with my mill. testing pcbs is so boring ;-)
<DocScrutinizer>
anyway doing this N^2 times for a unknown PCB with N (>1000) pads is a nogo
<DocScrutinizer>
having a way to sense all pads concurrently (like with my EL tester) will reduce the effort from N^2 to N
<wpwrak>
another way: heat the pad and look at the board with an IR camera
<DocScrutinizer>
there are possible optimisations for the N^2 scheme, by using combs with a number of 5mm brushes, then pad pintesting only the 2*5mm^2 rectangles where you got a signal
<DocScrutinizer>
(heat) wont't fly
<wpwrak>
(heat) would be worth a try ... if i had one of these expensive ir cameras :)
<DocScrutinizer>
you can't follow a trace >10mm in layer 4/8 this way
<wpwrak>
you should be able to see where it comes out again
<DocScrutinizer>
esp the fine ones used today, which are like 0.05mil
<DocScrutinizer>
nah
<wpwrak>
Cu should still conduct the heat a lot better than the FR4
<wolfspraul>
DocScrutinizer: why is this anti-anything? that's great stuff
<DocScrutinizer>
if it doesn't come out <10mm from your heated pad, you get nothing useful
<wolfspraul>
thanks a lot for sharing
<DocScrutinizer>
wolfspraul: it's about RE of closed hw :-)
<wolfspraul>
and analysis can be used for many different applications
<DocScrutinizer>
indeed
<wolfspraul>
I'm serious. not trying to 'hide' reverse engineering
<wolfspraul>
reverse engineering is an awesome technique in and of itself
<DocScrutinizer>
and the project itself can be attributed open hw anyway
<wolfspraul>
efficient analysis ideas and tools can come in handy in many situations
<wolfspraul>
so - great stuff! :-)
<DocScrutinizer>
this method, while not fit for decent QA of PCB, could come in handy for a quick and dirty good/bad test
<wolfspraul>
Adam is trying to explain to me for ages how important some particular pcb tests are, but I haven't found the time and concentration yet to mentally absorb it :-)
<DocScrutinizer>
as it detects one of the most annoying failure patterns of PCB: shorts
<DocScrutinizer>
I think when droned in clear water (with a bit of ions aka salt), and then applied some 10V to one pad and looking with a good camera for bubbles building up on other pads could actually work great
<DocScrutinizer>
I hope this chan is logged ;-D
<DocScrutinizer>
with proper date stamps
<DocScrutinizer>
you should probably make sure your DUT PCB has the hydrogen side of current flow, not the oxigen one ;-D
<wpwrak>
DocScrutinizer: try it and send pictures ?
<DocScrutinizer>
might consider it, yeah. Alas I have no good camera
<DocScrutinizer>
esp for the bga pads that are microscopic
<DocScrutinizer>
you'd need a really really good camera for that
<DocScrutinizer>
plus professional lighting, stand etc
<DocScrutinizer>
sounds like sth I could do for a master thesis :-D
<wolfspraul>
DocScrutinizer: yes the channel is logged and the logs were even beautified recently...
<DocScrutinizer>
nah, I bet this is like all my inventions: somebody did it 6 weeks ago, or will do in two weeks with no apparent link to this conversation
<DocScrutinizer>
and I'm not really in the situation for registering patents
<wpwrak>
DocScrutinizer: so, do it, take pictures, post it on slashdot. then it will be very hard for someone else to claim prior art later :)
<DocScrutinizer>
wpwrak: (heat) on vias you lose
<DocScrutinizer>
I bet a beer it's not worth trying
<DocScrutinizer>
what you *can* get by checking heat is the actual routing of a burried trace, when you send current thru it
<wpwrak>
(heat) dunno. they should still conduct the heat way better than the FR4. if you "inject", say, 100 K, even remote branches shuold get a delta of a few K (delta between heat conducted through direct trace vs. heat conducted through other traces or board material)
<DocScrutinizer>
but that'S usually not that much interesting (though I might recycle this isea for a special problem where I need to drill 2 holes into a PCBA)
<wpwrak>
yeah, current always works :) O(n^2), though.
<wpwrak>
are you trying to reverse-engineer something specific ?
<DocScrutinizer>
(heat conducting) the heat conducted to a far away pad directly linked by a trace is way less than that conducted to a parallel trace in a different layer, even when that other trace is thermally "isolated" by some plastic+glass PCB base material
<DocScrutinizer>
wpwrak: nothing special. Just looking at this N900 bare board and pondering about the few testpoints not mentioned in the 'leaked' schematics
<DocScrutinizer>
rhen recalling how I started my chitchat with OM, by ranting "gimme schematics, or I'll do <see above>" ;-D
<wpwrak>
;-)
<wolfspraul>
you can use sandpaper to go through the layers and then scan them
<DocScrutinizer>
indeed, but that's a lousy method, and frequently introduces lots of errors on vias
<DocScrutinizer>
and usually will need >5 PCB to ruin that way
<DocScrutinizer>
also takes "ages"
<wolfspraul>
I haven't done it myself, but definitely you can do it with one pcb
<wolfspraul>
"I used corning 120, and 400 for polishing it before scanning"
<wolfspraul>
"I used 120 really most of the time, and from one layer to another I would say something like 25 minutes, excluding the scanning"
<wolfspraul>
"400 is only used like 30-50 seconds :)"
<DocScrutinizer>
duh, nice job
<wolfspraul>
he needed maybe 1-2 boards to become skilled at this, but I think now he feels quite confident with it, and would only need 1 pcb to go through
<DocScrutinizer>
didn't think it could be done that cleanly
<wolfspraul>
here in China I can give boards to any pcb maker and they will have a kid sandpaper through it and return the scanned jpegs to me
<roh>
re
<DocScrutinizer>
lo roh
<wolfspraul>
DocScrutinizer: those pics are from China, I don't have a url for the boards steve|m sanded, but it also looked quite clean.
<wolfspraul>
it is slow and tedious, no doubt
<wolfspraul>
many hours with the sandpaper
<DocScrutinizer>
and won't show micro-short defects etc
<wolfspraul>
and you need a few iterations before you are good at it
<wolfspraul>
but other than that it's a simple and effective technique to picture the layers
<wolfspraul>
wpwrak: is it 'wow' because so low or so high?
<wpwrak>
so high ! isn't that what fully grown factory workers make ?
<wolfspraul>
yes but they just take one of their workers
<wolfspraul>
"today, you have to sandpaper this pcb"
<wolfspraul>
:-)
<wpwrak>
great way to escape the daily boredom of the factory line ;-)
<roh>
i wonder how great quality and ingenuity of chinese products could be if they had such good teachers as some of us had here in germany
<roh>
teachers of professions.. like if you end up in siemens 'lehrwerkstatt' for some time
<roh>
not neccessarily a nice experience, but usually one which makes your handywork improve a lot
<dvdk>
why does package plplot cause openwrt to compile octave, too?  didn't even select sub-package plplot-octave.  or aren't sub-package dependcies not tracked independently?
<kristianpaul>
wonder what you can buy with 500usd in china
<whitequark>
does anyone have docs on Ingenc SIMD instruction set?
<kristianpaul>
nope :(
<kristianpaul>
but
<kristianpaul>
dingux wiki have some data about it
<kristianpaul>
i heard also mplayer source code from ingenic have it
<kristianpaul>
have imoplemented SIMD as asm,
<whitequark>
kristianpaul: the mplayer code is a huge load of crazy shit
<whitequark>
you'll need to be an ingenic developer to understand even a small bit of it
<whitequark>
the dingux wiki has a really great manual
<whitequark>
thanks
<kristianpaul>
yeah but not said i dint point you to mplayer
<whitequark>
kristianpaul: well, I knew about mplayer and already checked it as a possible source
<whitequark>
and it's an awful source of information for a complex thing like SIMD ISA
<kristianpaul>
btw can you tell us what are you planning to do with SIMD and the nanonote?
<zear>
hey
<kristianpaul>
hai
<zear>
i left you guys just for a moment and you've done a wireless adapter already? :D
<kristianpaul>
pcb so far
<zear>
well, as long as it works, i'm gonna order two :D
<kristianpaul>
SMT is a WIP but not updates from tuxbrain tha i'm aware off
<kristianpaul>
yeah it does zear
<kristianpaul>
did you saw wpwrak maill about the tunnel he made to do ssh and ping between nanonotes?
<zear>
i think i've seen the vid
<kristianpaul>
ya, that famous video indeed ;-)
<zear>
btw, did owrt got any better in the terms of the libs and toolchain?
<zear>
so i can port some games without a bigger hassle?
<kristianpaul>
well. i guess you should consult about owrt libraries avaliabillity
<kristianpaul>
about toolchain i cant talk too much about it, as the only cross compile i had done for the nanonote was using jlime :p
<kristianpaul>
so wait xiangfu or ask kyak  i hink
<kristianpaul>
think*
<zear>
yeah, same, with the exception i also used the dingoo toolchain with statically linked binaries :D
<kristianpaul>
hey, i about to point you that (statically linked games)
<kristianpaul>
i cant see better use for something static, well including kernel may be, and btw does nanonote kernel from owrt suppport dynamic module loading?
<zear>
kristianpaul, you could like the binary dynamically and keep your own libs in the local directory
<zear>
and change the priority of libs over the system ones
<kristianpaul>
oh
<zear>
export LD_LIBRARY_PATH or something
<kristianpaul>
ah, true :-) i forgot is all about those env variables..
<kristianpaul>
so, what games are you wishing to port zear ? :o
<zear>
kristianpaul, i already did some ports very early in the nn history
<zear>
yet before the first batch
<zear>
though it was all statically linked with the dingoo libs
<zear>
i did spout, prboom, eduke32
<zear>
also cave story, which i never released
<zear>
brb
<kristianpaul>
ah yes i remenber prboom
<kristianpaul>
and gmenu2x wich not a game but very important
<kristianpaul>
and that make me remenber somebody was asking for dmenu the other day
<Jay7>
one day someone was asked me about making standalone fb menu from kexecboot :)
<kristianpaul>
Jay7: yes :-D
<wpwrak>
Jay7: makes perfect sense. ideally, kexecboot should just be a bundle of general components. the fewer kexecboot-specific parts you need, the better.
<Jay7>
btw, may be good idea to do this..
<Jay7>
fb lib + menu lib + xpm lib + ... :)
<kristianpaul>
yay, the ENC28J60 just get to my mother work :-)
<kristianpaul>
i'll get some fun with linux soon it seems
<kristianpaul>
wpwrak: if it is chip compatible i can order one why not
<kristianpaul>
so far this is just 10Mb/s
<wpwrak>
the chip operates with 3.3 V, so it should be fine
<wpwrak>
seems to need 5 I/O lines. UBB has 6. more than enough ;-)
<kristianpaul>
zear: an nc...
<kristianpaul>
well i was in thinking something else
<kristianpaul>
like a kexecboot-like menu :p
<zear>
kristianpaul, this thing can actually launch binaries
<kristianpaul>
zear: yeah but is a file browser..
<zear>
well, i don't really see a difference ;)
<kristianpaul>
well,if your apps are in the same dir, will look like more a menu.
<zear>
yep
<kristianpaul>
zear: have you seen kexecboot menu?
<zear>
looking at it now
<zear>
looks nice, but a file browser is so much easier in use ;)
<kristianpaul>
really?
<zear>
i can only imagine the ordeal of adding new links to that thing :)
<kristianpaul>
well, thats a task for the developero, but if the menu worsk, load fast and the know-workign apps get lunched and all wokrs like a charm... well :-)
<kristianpaul>
also i forgot to said, somebody was/is digging and removing som unusefull code from gmenu2x, thats could change things
<whitequark>
kristianpaul: is the source code for that fork published?
<whitequark>
that's interesting
<kristianpaul>
may be last image already included that.. well i dont know
<wpwrak>
zear: a file browser is also a little fat ...
<kristianpaul>
whitequark: in projects.qi-hardware.com there is one for gmenu2x fork i think
<kristianpaul>
i know the code is around
<kristianpaul>
i saw several commits from qi-bot last weeks
<zear>
wpwrak, i thought you guys were a bunch of geeks who dislike anything that's fool proof ;)
<kristianpaul>
no no, well thats not the idea i think
<zear>
well, gmenu2x does the job just fine, doesn't it?