#milkymist IRC log for Sunday, 2011-04-03

kristianpauldont think so (720p)00:03
kristianpaullars_: got boot linnux on mm1?00:05
lars_sort of00:27
lars_both qemu and the real hw crash as soon as the kernel tries to execute a userspace application00:32
mwallebFLT format?00:33
lars_hm, no. although i'm using the elf2flt ld script00:38
lars_but for some reason the resulting executable is still elf00:39
mwallethere should be two executables00:39
rohlekernel: you mean 720p in analog? not so much. if you can do sync on one of the signals and yuf colorspace then maybe one can use that00:41
rohwith a vga to 3x rca cable00:42
lars_mwalle: file busybox00:45
lars_busybox: BFLT executable - version 4 gotpic00:45
lars_"BINFMT_FLAT: reference 0x657272 to shared library 32, killing sh!00:47
mwallelars_: shared libraries?00:47
mwallethere is no support for shared libs yet00:47
mwalleonly static linked bFLT binaries are working00:48
mwalleneed to go to bed now, gn800:48
lars_well the elf binary is build as static binary00:56
CIA-43flickernoise: Sebastien Bourdeauducq master * r2720bdd / src/filemanager.c : File manager: clear, rename and mkdir commands - http://bit.ly/f8pQ9w10:11
CIA-43flickernoise: Sebastien Bourdeauducq master * re1e722f / src/filemanager.c : File manager: delete command - http://bit.ly/gyzMHq10:58
mwallelekernel: philpem asked me to change the jtag cores license to a more permissive one. so here it is11:00
CIA-43milkymist: Michael Walle master * r66ffb15 / (2 files): change license for lm32 jtag cores - http://bit.ly/iiWzou11:03
kristianpaulnew bsd ! :)11:09
wpwraklekernel: did you take the screenshots for the new color theme in qemu or already with fbgrab ?11:49
larscmwalle: could it be that the symbol address calculation is wrong?11:56
larscall the relocations are of type LM32_32, but the data at sym_addr doesn't really look like an address that needs to be relocated12:44
mwallelarsc: how does it looks like?13:07
mwallewhat does flthdr -p say?13:07
mwalledid you try the -a parameter for elf2flt?13:07
mwallelekernel: why not?13:08
larscmwalle: http://pastebin.com/HHpzSVrw13:14
larsccould it be that has-pic-got is causing trouble?13:17
mwalleyes, that flag should not be set13:18
mwalleare you passing the -p parameter to elf2flt?13:19
lekernelwpwrak: with fbgrab13:20
larscmwalle: i use ld-elf2flt13:21
lekernelactually, qemu seems to break when you try to switch to 1024x768... i'll have a look at it13:21
lekernelmwalle: well, it's up to you :) as long as it's compatible with the GPL, I don't care :)13:22
mwalleiirc my linker parameter was -Wl,-elf2flt or -Wl,-elf2flt=-a13:23
wpwraklekernel: (fbgrab) kewl. so we already have a first "real-life" usage example.13:23
lekernelwhat a bunch of crap http://www.hdmi.org/manufacturer/terms.aspx13:27
lekernelanyone already tested the "we did not sign anything" technique, like with the memory cards?13:28
mwallelekernel: are you using 4 or 8 spaces per tabstop?13:28
lekernelmwalle: 813:28
larscif i manually disable -p i get 'bad realoc 11,12,13'13:29
mwallelarsc: could you please send me the elf binary?13:39
mwallebad_relocs occur when there is a unsupported reloc type13:40
larscmwalle: http://metafoo.de/busybox_unstripped.gdb13:41
larsc11,12,13 are all GOT related13:41
wolfspraullekernel: I'd say it's as usual: 1) trade secrets 2) patents 3) copyrights 4) trademarks13:43
wolfspraulwe can easily work without using their trademarks, that's an easy exercise13:44
wolfspraultrade secrets are only trade secrets as long as they are a secret, so if you find specs somewhere, it's not a secret anymore13:44
wolfspraulcopyrights - well, we all know how this works, and of course we cannot use copyrighted material13:44
wolfspraulpatents - that's an open ended one, we can safely assume in our case it should be mostly FUD, but you never know it would require a bit more investigation13:45
wolfspraulit's definitely a difficult subject. Best might be to even approach them openly about our clean-room implementation...13:46
wolfspraulif you get the feeling they are super aggressive, you may not want to go there even if you think the law is on your side13:46
mwallelarsc: whats your busybox config?13:47
wolfspraulI know very little about HDMI though, maybe we should research a bit.13:47
lekernelwolfspraul: the same protocol is used for DVI13:47
wolfspraul'HDMI' is definitely a trademark already, I would think.13:47
lekerneland the standard (for DVI) is publicly accessible13:48
wolfspraulthey will not help you understand which of their generous services you do not need :-)13:48
wolfspraulthat work you have to do yourself...13:48
wolfspraulwe need to stay away from trademarks13:48
lekernelwhich has enough information to make a FPGA core able to display pictures on HDMI displays13:49
wolfsprauland from copyrighted material that we have no license for13:49
wolfsprauland we should not do anything illegal to break into a trade secret13:49
wolfsprauland we should respect/work around patents as much as that is possible13:49
wolfspraulall obvious of course...13:49
wolfspraulno news13:49
wolfspraulwell great then, seems you have a start already13:49
mwallelarsc: http://paste.debian.net/112846/13:49
lekernelit's actually not very difficult to do13:50
lekerneland regarding hardware it's merely adding the connector and wiring it up to the FPGA pins (the spartan6 has built in compatible transceivers already)13:50
wolfsprauljust stay away from their trademarks, this time from the beginning13:51
larscmwalle: the same here13:51
wolfspraulthey may very well not add much more than their trademark to an otherwise open spec, but if they do that it is still their trademark13:51
wolfspraulif they managed to create some value (recognition) with this trademark, then so be it13:52
wolfspraulwe cannot use it13:52
lekernelbut well, later. I don't even have an HDMI display to test with13:52
lekernelbut if at some point we want to add it, it shouldn't be a big problem13:53
lekernelwe can call it "HD connector" or something like that13:53
wolfspraulHD = high definition?13:53
wolfspraulwhy 'definition'?13:54
lekernelwe'd rather not spend those 10k annual license fees plus per-device royalties... better use that money elsewhere13:54
wolfspraulwhat's the difference between VGA and HDMI? HDMI is digital, right?13:54
wolfspraul'digital video output'13:54
lekernelyeah :)13:54
lekernelcould work13:55
lekernelI'm not sure about the trademark status of "HD"13:55
wolfspraulbut I would err on the safe side13:55
lekernelthere is this "HD ready" logo which is definitely trademarked - and the license, on top of being expensive, requires that you implement DRM13:56
wolfspraulstay away from 'HD', sounds like13:56
wolfspraulsame as 'SD'13:56
wolfspraulguess the 'D' is attractive :-)13:56
wolfspraulwe really should not piggypack on a trademark, it will backfire, for sure13:57
wolfspraulif they created value/recognition with this 'thing', whatever it is, then it's theirs13:57
wolfspraulI'm in China where people trample over trademarks all over and I know that is no fun.13:57
wolfspraulNokic, Nckia, Samsnug, Smasung, Sumsung, and so on13:58
mwalleis DVI a trademark too?13:58
mwallelekernel: my dm800 has a dvi connector with a dvi-to-hdmi cable14:02
wolfspraulWikipedia says "DisplayPort has an advantage over HDMI in that it is royalty-free, while the HDMI royalty is $0.04 USD per device and has an annual fee of $10,000 for high-volume manufacturers."14:05
lekernelthe DVI connector is rather big14:05
lekernelbtw this "HD ready" logo thing is rather evil. make the consumers want "HD ready" products, then require DRM to be implemented to license the trademark14:06
mwalleminidvi? :)14:06
mwallewell apple crap14:07
lekernelnow I can troll the random geek who wants HD by pointing out that they also mean they want DRM14:07
wolfspraullekernel: yes sure, but remember they think they create 'value' for all sides14:07
wolfspraula way to pay is a way to create value, create a product14:08
lekernelI think the HDMI connector is quite good... small and compatible with many devices14:08
wolfspraulso that's a good thing in their thinking, understandably imho14:08
lekernelthere are also HDMI->DVI adapters14:08
lekernel(it's the same signaling)14:08
lekernelif we can use that connector without too much legal trouble, good14:09
wolfspraulat first I would say we can.14:09
wolfspraulunless they have some really evil patents on mechanical hooks in the connector.14:09
lekernelhttp://en.wikipedia.org/wiki/Digital_Visual_Interface says "superseded by HDMI"14:10
wolfspraulunfortunately most likely they will do little to educate us - that falls 100% on us14:10
wolfspraulI'm not worried about this stuff, let's just be careful about the points I outlined above, as usual.14:10
wolfspraultrademarks to begin with14:10
mwallelarsc: mh i dont have any got symbols in my elf binary14:15
mwallelekernel: btw you could append "pld reconfigure" to the flashall.batch script14:21
kristianpaulwhat is reconfigure for?14:34
mwallekristianpaul: forces a fpga reconfiguration, eg. loading its configuration again (from strapped source, that is nor flash)14:48
kristianpaulmwalle: oh, nice i tought that dint existed, you save me some powercycles now :-)15:17
scrtshm, does milkymist use RTP/RTSP?15:18
kristianpaulyou mean for video streaming?15:21
kristianpaulyou'll need to implement it afaik15:25
scrtshm, so the ethernet is used only for control and use udp?15:28
kristianpaultcp too15:30
scrtshm, okay :)15:31
kristianpaullekernel: do you have somwhere the flickernoise version with support 1024x768?16:22
kristianpaulhmm, but i still need check if the vide projector support that resolution16:23
kristianpauli mean if you have a binary :-)16:23
Fallenouscrts: the ethernet is a generic ethernet, you can do whatever you want with it, at the moment it is used for telnet, ftp, nfs and netboot16:32
scrtsheh, well I am interested how RTP packet is formed using no processor, pure HDL only16:33
Fallenoulayer 7 via fpga is quite rare :)16:33
rohFallenou: nope. not for rtp16:34
FallenouI don't know rtp in details but as far as i know it's just streaming payload + timestamps ?16:34
rohneed to run now.16:34
Fallenouit seems doable16:34
rohit is. seen it.16:35
Fallenouoh yes16:35
Fallenouthanks for the link16:35
scrtshmm, well pretty fast if there are more than 1 video stream :)16:50
mwalledoes anybody know if gdb, the stub or the hardware is supposed to increment the PC in case of a watchpoint trap?17:27
CIA-43milkymist: Sebastien Bourdeauducq master * r1fe5d20 / boards/milkymist-one/flash/flashall.batch : Reload FPGA after flashing - http://bit.ly/i3ly7g18:11
lekernelFallenou: http://www.youtube.com/watch?v=7syu5EC1OWg18:48
mwalleomg :)19:01
lekernel##[|#æ! rtems is such a pain at times19:06
lekernelrename() won't support cross-mountpoint moves19:07
lekernelhow I love wasting my time reinventing such wheels19:08
tuxbrainlekernel:(fpga web server) wow, can this be implemented on milkimist in adition of what it has now?19:08
lekernelof course, but it would be rather irrelevant to do so19:08
lekernelthat's what CPUs are for19:08
lekernelunless you need very high performance (and still...)19:09
tuxbrainmaybe to us it as front end without whaste cpu cicles than may affect to the VJ function19:10
tuxbrainjust talking from the most complet ignorance as you know19:10
lekernelcontrary to its filesystem, the RTEMS real time scheduler is pretty nice and can handle such things19:10
lekerneland there's also the option of making a dual-CPU system, which would still be better than a HW webserver19:11
tuxbrain:( www.gawgle.net doesn't work19:11
tuxbrainno more info about it :(19:12
tuxbrainI wanna see if it has any server side scripting capabilities like php/phython/perl...19:13
lekernelhahaha, most probably not19:14
lekernelit's probably a big hacky FSM19:14
tuxbrainFlying spaguetti monster???19:14
lekernelfinite state machine19:14
tuxbrainok :P19:14
Action: tuxbrain dreams on dinamic webservers on a chip19:15
tuxbrainimagin that to integrate along wpan on tinny low power modules19:17
tuxbrainall devices in a home from the ringbell to the alarm clock can be setup and consulted using a web interface wiresly, even yous cloths can be part of the network ....19:20
mwalleso use some low power uC :)19:27
rohuc isnt the issue. the rf eats 100 times more20:13
wpwrakroh: depends on your transmit power :)20:14
rohtuxbrain: if you need a 'cpu and transciever board' .. check out the chibiduino20:19
roh33$/piece is quite cheap. i dont think one can easily beat that20:19
tuxbrainroh I have an eye on chibiduino, I think once NN has also wpan could be a great companion to interact wiresly with the real wold :), also I hope MM would have an driver for atusb soon to interact with chibi or NN :)20:23
wpwraktuxbrain: just install linux on the MM ;-)20:26
tuxbrainhe :), wpwrak you know I will love to see a full trootle linux on it I know some people are improving the actual port, or at least willing to do in short, but I would like to see atusb running on the "official" rterms version20:28
tuxbrainout of the box MM NN interaction :)20:28
Fallenoulekernel: already seen this video :)20:29
Fallenounice but useless :p20:29
tuxbrainmaybe an app with a front end on NN to change patches/edit/create patches on MM wiresly20:30
lekernelroh: when you see the price of that hot wheels gun, *duino looks expensive20:31
lekernelbut yeah, for small batches, $33 is cheap20:31
rohlekernel: it was also 20E or so last time i checked20:31
lekernelsure, but it has case, mechanics, waveguide, AVR, LCD, batteries, and that microwave module which are typically in the 200E range instead of 2020:32
rohmicrowave stuff isnt expensive. its just some metal anyhow20:32
roha lnb is more complex and also <10E20:33
lekernelyeah. see the power of mass production20:33
lekernelif you tried to build a LNB yourself (with regular PCB and chips), you'd pay a LOT more20:34
rohsure. but because of development, not material cost20:34
lekernelno, for MW it's often material cost20:35
rohits mostly a combination of modern semiconductors and quite simple rf circuits with extreme focus on mechanical layout20:35
lekernel#!@§ rtems filesystem bugs (?)20:40
lekernelseriously the rtems filesystem is the shittiest piece of code i've discovered in the past 6 months (autocrap doesn't count, I knew it before)20:41
lekernelthere's that utterly retarded eval_path system, and open file then delete file from another task = crash20:43
lekernelnow it seems that the shell commands (rm/mv/cp) fail intermittently when run from the app and not the shell. and my guess is this might be related to eval_path bugs ... :(20:45
lekernelwow, I even managed to create an undeletable + unreadable directory on the ramdisk using cp alone20:46
Fallenouyes eval_path is a shitty thing20:49
Fallenouit's not even using some sort of vfs, is it ?20:49
Fallenouthey really should be using a vfs kind of thing20:49
lekernelhaha, no, eval_path replaces the vfs20:51
lekernelit's kind of a VFS name resolution function split across every registered filesystem. it's a bit as if Salomon carried on with his idea of splitting the baby, with blood spilling everywhere20:54
CIA-43flickernoise: Sebastien Bourdeauducq master * r95168a6 / src/filemanager.c : Filemanager: complete - http://bit.ly/fhMcE021:23
CIA-43flickernoise: Sebastien Bourdeauducq master * rd2256c5 / (6 files): More DMX channels - http://bit.ly/iaITwm21:41
CIA-43flickernoise: Sebastien Bourdeauducq master * r5f9a047 / flash/flash.sh : Reconfigure FPGA after flashing - http://bit.ly/dEEojo22:01
CIA-43mtk: Sebastien Bourdeauducq master * ra478a85 / (5 files): Removed view manager - http://bit.ly/fWYSeO22:24
lekernel  CC      drivers/video/milkymistfb.o22:55
lekerneldrivers/video/milkymistfb.c: In function 'milkymistfb_probe':22:55
lekerneldrivers/video/milkymistfb.c:553:2: error: implicit declaration of function 'out_be32'22:55
lekernelmake[2]: *** [drivers/video/milkymistfb.o] Error 122:55
larscreplace it with iowrite3223:05
lekernelyup. done23:06
lekernelnow I get that gcc infinite loop in timer_list.c again... should "optimize for size" iirc23:06
lekernelok, works :)23:26
--- Mon Apr 4 201100:00

Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!