#milkymist IRC log for Saturday, 2011-04-02

xiangfulekernel: small patch for input.c http://pastebin.com/eTXxcJsu02:09
xiangfuI found this:02:09
xiangfuguirender.c:mtk_input(&e[i+1], count-i-1);02:09
xiangfuyes. it not cause any problem. just fyi :)02:10
xiangfulekernel: the shortcut for 'fbgrab' patches. sended out06:21
xiangfuusb cable for daughter board http://ntcdistributing.com/images/UH2-2MBR02-12large.jpg06:49
xiangfuwolfspraul: I think this usb cable is not good for milkymist one06:50
xiangfuI only remove the 'Video-in Side' keeps all other. and screw the top.06:51
wolfspraulxiangfu: yes I know07:01
wolfspraulI think for real developers who always, for months, want access to jtag, they will all do that07:01
wolfspraulbut then there are people who need jtag only once every few months, maybe for a major update. At least until we have the GUI update done really well.07:02
wolfsprauland for those it is easier to just take off the top07:02
wolfspraulthen reflash, then put the top back on07:02
xiangfuoh. yes. that make sense.07:02
wolfspraulbut if you constantly want jtag, you will instead take the side element out, and put the top back on07:02
wolfspraulthen you will need a normal USB cable07:02
wolfspraulgetting the side element out is quite hard, you have to remove the screws of the video-in connector, and maybe also the screws on the bottom so that the side part is loose enough to be lifted07:03
wolfspraullet's see how all this goes.07:03
wolfspraulxiangfu: but I totally agree with your comment.07:04
wolfspraulI will also leave the side open.07:04
wolfspraulsome people may also add an extra hole on the side07:04
xiangfuI am think screw a hole on my 'video-in' side element :)07:04
wolfspraulyes :-)07:04
Action: kristianpaul just make a hole for the jtag cable,of course using proper tools07:04
wolfspraulthe upward pointing USB cable has limited value, for sure07:04
wolfsprauland the limited value may even go lower in a few months as we improve everything07:05
wolfspraulso we may drop it again at some point. maybe we even drop the entire jtag-serial board at some point. Don't know.07:05
wolfspraulnow we need these things, to make the starting point easier, for a lot of people in different use cases.07:05
wolfspraulI doubt the news have a big impact, but we need to keep the routine and keep pushing for a wider audience...07:06
wolfsprauloops, wrong channel :-)07:06
xiangfukristianpaul: don't have that tools for now :)07:08
lekernelwolfspraul: what's wrong with the current gui update?08:59
lekernel"At least until we have the GUI update done really well."08:59
lekernelxiangfu: why do you need "[PATCH 1/3] delete input callbacks from last"?09:01
xiangfuhmm.. I found the main.c and guirender.c both run "input_add_callback(mtk_input);" and guirender.c run "input_delete_callback(mtk_input);" every time it delete . it delete the mtk_input in main.c09:04
xiangfulekernel: sorry. maybe I man wrong. just ignore that patch.09:04
xiangfulekernel: I should look into more detail.09:05
xiangfujust ignore that patch.09:05
lekernelxiangfu: what guirender does is disconnect MTK from the events when rendering09:06
lekernelso MTK does not draw stuff to the framebuffer anymore09:06
lekernelotherwise you'd see flickering buttons, mouse cursor etc. when touching the mouse during rendering :)09:06
lekernelit's a bit hacky, but it works09:07
xiangfulekernel: thanks for the info.09:18
xiangfusome question. there is a "stop_callback" in guirender.c09:18
xiangfuif there any reason the the stop_callback have to run inside  guirender.c when the guirender() success execute09:20
xiangfuthe rename the reboot.c to shortcuts.c. and add the "CTRL + F12" to that file.09:23
lekernelyeah you can move it to reboot.c09:24
lekernelso you have only one event hook09:24
lekernelthe stop_callback in guirender.c has nothing to do with the event hooks09:25
xiangfulekernel: ok.09:28
xiangfulekernel: I think I will try to focus on mkyaffs2image next few days.09:28
xiangfumy TODO list on milkymist is here: http://en.qi-hardware.com/wiki/User:Xiangfu#Milkymist_One09:30
xiangfuany feedback will be great :)09:30
xiangfu(btw I test "creating a sub-folder on the flash and then uploading a file via FTP",test 24KB file and 3MB file. 24kb file works fine. 3MB file have problem. the m1 just freeze)09:31
xiangfuback online later. will read log :)09:37
wolfspraullekernel: maybe it is evolving quicker than I can catch up. What's wrong is that I haven't tried yet :-)10:17
wolfspraulso I should rephrase "after Wolfgang realized how good the GUI update already is, ..."10:17
lekernelthere aren't many issues with the GUI update. it's just select file and click flash.10:18
lekerneland I haven't found a single bug there so far10:18
wolfspraulok, but I also want it to go to a website and download stuff10:18
lekernelotoh if you're using FTP to transfer the files you might run into the ethernet bugs10:18
wolfspraulwill xiangfu's fbgrab also work during rendering?10:20
wolfspraulwill it stop rendering for x seconds?10:20
wolfspraulok but it works, great10:22
CIA-43flickernoise: Xiangfu Liu master * r17de0a3 / (5 files):11:05
CIA-43flickernoise: move all fbgrab to one module11:05
CIA-43flickernoise: Signed-off-by: Xiangfu Liu <xiangfu@sharism.cc> - http://bit.ly/dJIZjH11:05
CIA-43flickernoise: Sebastien Bourdeauducq master * r3fe7024 / src/Makefile : add fbgrab to Makefile - http://bit.ly/hHscbq11:05
CIA-43flickernoise: Xiangfu Liu master * r953d9bb / (5 files):11:06
CIA-43flickernoise: add global shortcuts module bind Ctrl + F12 to 'fbgrab'11:06
CIA-43flickernoise: Signed-off-by: Xiangfu Liu <xiangfu@sharism.cc> - http://bit.ly/ehQGrL11:06
CIA-43flickernoise: Sebastien Bourdeauducq master * rbda6c50 / src/Makefile : update Makefile - http://bit.ly/hQDKoo11:06
wolfspraullekernel: in the email to rejon, what do you mean with "proper lighting and black background"?12:45
wolfspraulyou mean black background for the projector to project to? light what?12:46
lekerneluse a black surface behind the dancer/object so the background doesn't get into the visual effect12:46
lekernelcamera+light | dancer | black background12:46
wolfspraulah I see. both refer to video-in.12:46
wolfspraulgot it12:47
lekernelI use a piece of black cloth I can hang when using it12:48
wolfspraulI never tried the actual video-in rendering, only made it to the preview dialog so far :-)12:53
xiangfukristianpaul: lekernel the IRCLOG search fixed. :)  http://en.qi-hardware.com/mmlogs/search?q=testing15:16
xiangfuthere are one typo in apache  configure file.15:17
kristianpaulhmm ;)15:17
xiangfukristianpaul: but the log file you send to me have early logs which en.qi-hardwware.com/mmlog don't have :)15:19
xiangfukristianpaul: "/path" typo.15:19
kristianpaulxiangfu: yeah :-)15:20
lars_'shell/ash.c:70:3: error: #error "Do not even bother, ash will not run on NOMMU machine"' :/20:50
lekernelyeah stupid20:55
lekerneland I'd even say never mind :)20:56
lekernelso I guess the linux port is doing pretty well, now that you're trying to compile new apps?20:56
lars_i finaly found time today to put everything together20:58
lars_hm... of course it does not work on the real hw :/22:53
lekernellars_: works in QEMU and not on real hw? what happens?23:16
lekernelbtw is the instruction cache properly flushed when dynamically loading code?23:17
lars_lekernel: the kernel doesn't show any life signs23:21
lekernelhuh... it worked before23:22
lars_i'm trying to bisect it23:22
lekernelwhat happened?23:22
lars_no idea23:22
lars_i switched from .37 to .38 and added mwalles patches23:22
lars_and it still works in qemu23:22
lekernelFallenou: does RTEMS support M_EXT for mbufs?23:28
lekernelbecause if it does, you should not need to copy memory in the ethernet driver23:28
Fallenouwill check23:29
lekernelthere's also this cluster thing which might allow a copy-free driver, but I don't remember how it works23:30
Fallenoulekernel: yes rtems does support M_EXT23:31
FallenouI will try to use it23:32
kristianpaulcopy-free driver :o23:32
lekernelmaybe the right thing to use is the "mbuf cluster"23:33
lekernelit should support up to 2048 byte packets23:33
lekernel m_pulldown(struct mbuf *m, int off, int len, int *offp)23:33
lekernel             Ensure that the data in the mbuf chain starting at off and ending23:35
lekernel             at off+len will be put in a continuous memory region.  len must23:35
lekernel             be smaller or equal than MCLBYTES.23:35
lekernelMCLBYTES == 204823:35
lekerneland MCLBYTES is the size of a mbuf "cluster" (whatever that means)23:35
Fallenouhum hum will se maybe tomorrow23:35
Action: lekernel hasn't used mbufs since 200523:35
Fallenoui'm sick and my head is starting to explode23:35
Fallenouwill goto bed23:35
Action: kristianpaul looks what is a mbuf23:35
Fallenoukristianpaul: a structure that contains ethernet packets, in bsd world23:35
Fallenouequivalent of skbuff of linux23:35
Fallenouto avoid too much copying23:35
kristianpauloh i see23:35
lekernelah, mbuf cluster is just a special case of M_EXT23:39
lekernelThe system also supplies a default type of external storage buffer called23:39
lekernel     an mbuf cluster.  Mbuf clusters can be allocated and configured with the23:39
lekernel     use of the MCLGET macro.23:39
lekernelI don't think we should use the mbuf cluster23:41
lekernelbut use MEXTADD() and recycle the data buffers (with proper locking ofc to avoid problems like the random crashes we see now)23:41
Action: lars_ blames mwalles patches. 2.6.38 without them works fine23:48
mwallehey :b23:49
mwallewhich patch?23:49
lars_don't know yet23:51
lars_i applied the whole series you sent some time ago23:51
mwallelars_: and the kernel doesnt print anything?23:53
lars_even with earlyprintk23:54
lars_hm, now it works23:57
lekernelbtw, is 720p over VGA common?23:59
--- Sun Apr 3 201100:00

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