#qi-hardware IRC log for Tuesday, 2013-10-08

DocScrutinizer05hah, nice00:05
DocScrutinizer05I don't think they will create that oled transparent00:06
wpwrakhmm, the middle button of my mouse seems to have died suddenly :( first time something like that happens08:13
viricyou need a mouse without mechanical parts08:14
wpwrakor with better mechanical parts08:17
wpwrakthat mouse wasn't even particularly old08:18
uncloudedmice seem to be going downhill recently.  I had fairly pricey Logitech RMB fail on me recently08:22
wpwrakthis one's a logitech, too. M215. i liked that it ran off a non-rechargeable battery.08:29
wpwraklogitech's web site is refreshingly honest: http://www.logitech.com/en-us/product/wireless-mouse-m185?crid=708:29
wpwraklots of bad reviews08:30
viricI just buy a laser mouse at the nearest shop08:42
viricif within two years it breaks, I return it08:43
viricand get one new for free08:43
uncloudednice shop.  I've also got a M905, which is so beautifully made that you'd hardly believe that it was made by the same people who put the appalling G500s that went wrong08:45
wpwraki guess i'll get another M200 +/- 15. the fancier models all may need a special setup on the pc.08:50
uncloudedthe M905 doesn't.  The unifying feature doesn't work on Linux apparently but just using as a normal USB mouse works fine08:54
uncloudedit does really work on glass tabletops and mirrors too08:54
wpwrakmy my desk's solid wood. any at no risk of getting too clean for the mouse to have something to recognize :)08:56
viricI always use a mouse carpet, with a textile-like surface.08:57
wpwrakhmm, have to add an n-FET to the card power switch. else, it'll go "always on" when the battery gets low enough that Vbat + the FET's threshold voltage < 3.3 V (+ margin)09:41
wpwrakwith the threshold as low as 0.45 V, that can happen relatively easily09:41
wpwrakit's good that there's plenty of room for this kind of rework under the OLED ;-)09:42
larscdoes anybody happen to know by chance what populates /dev/ on ubuntu these days? Is this done by upstart?14:17
larschm, not on my machine, strage14:23
viricdevtmpfs is very old14:24
viricudev does14:24
pcercueiudev requires devtmpfs14:25
viricok, you win.14:25
viricI've devtmpfs here, but not ubuntu.14:25
pcercueidevfs is old, devtmpfs is not14:25
viricnone on /dev type devtmpfs (rw,relatime,size=202716k,nr_inodes=497343,mode=755)14:25
viric200MB. looks like too much.14:25
pcercueilarsc: 14:27
pcercueicat /sys/kernel/uevent_helper14:27
pcercueion my Debian, it doesn't return anything (because it uses devtmpfs)14:28
pcercueimaybe it's upstart then14:29
larscbut I have no /sbin/hotplug on my system...14:29
pcercueiand /dev is tmpfs?14:29
pcercueiI have no idea what "realfs" is14:30
larscnon-tmp fs ;)14:30
larscthere is nothing special mounted at /dev14:31
pcercuei/sbin/hotplug might exist... inside the initramfs14:31
larschm, which I do not have14:32
pcercueiwhy do you use ubuntu anyway? :o14:32
larsclinaro ubuntu14:32
larscon one of the arm boards14:33
paul_boddieAnyone here have any experience with Linux cdc_adm and serial-over-USB?16:39
paul_boddieI saw that wpwrak had his name on some Openmoko software of that nature. ;-)16:39
larscI used it back in the openmoko days16:40
paul_boddieDid you ever experience problems with the device it created? Not the problem half the Internet has with modem-manager wanting to own it, but with the driver not apparently listening on the IN endpoint.16:41
Action: paul_boddie notes that Amarok on KDE 4 is such a joke these days. Let's not play music any more, then!16:43
wpwrakuse mocp :)16:44
paul_boddieIf KDE were a reality show, half of the applications would have been voted off in KDE season 4.16:45
paul_boddieWith regard to cdc_acm, I'm trying out some AVR USB stuff that times out when trying to send to an IN endpoint, and I wondered if there was some known issues with the cdc_acm driver.16:50
paul_boddies/IN endpoint/IN endpoint on the host/16:50
paul_boddies/was some/were some/16:51
DocScrutinizer05qi-bot, the better edlin16:52
paul_boddieYes, the Internet can still learn from the "old school" tools.16:53
paul_boddieDocScrutinizer05: Was entertained/appalled by some of the messages on the Neo900 Maemo talk thread.16:54
paul_boddieThe mental model of how hardware gets to the shops seems to involve pixies and unicorns for some people. ;-)16:55
paul_boddieOr a secret Nokia factory still producing the devices in stealth mode and giving them away because they aren't allowed to sell them any more.16:57
paul_boddieBeing forced to do so by an armed unicorn.16:57
DocScrutinizer05the same unicorn that fires his shrink-gun at the chips16:58
larscpaul_boddie: never had any problems with the driver16:59
paul_boddielarsc: OK. Thanks for the negative result. ;-)16:59
larscit's probably some kind of combination between avr and driver17:01
paul_boddieYes, reading the driver source code can be a real eye-opener sometimes.17:02
paul_boddieThe device gets created but the AVR thinks its FIFO isn't ready for sending. The weird thing is that I tried an example using HID events and that works.17:03
Action: paul_boddie ends up learning useless things about USB again, this time for AVR...17:04
larscso you are in control of the AVR code?17:05
paul_boddieYes. I have been playing with it, even removing the CDC stuff and trying to make it behave like a vanilla serial device (using usbserial instead).17:06
paul_boddieIs there anything special I have to do when using these /dev/ttyACM* or /dev/ttyUSB* things?17:07
paul_boddieWith the ACM stuff I've seen stty being used, which is a real blast from the past.17:08
larscI actually had a problem with a cypress uart chip and the ACM driver17:08
larscthe device would show up17:08
larscbut never receive anything17:08
larscmaybe the same issue17:08
larsccypress solved it with an update to their chip17:08
larscwith an firmware update17:09
paul_boddieThere are a few things that people seem to have problems with, but they aren't showstoppers: the "not a modem" complaint from the driver, and mtp-probe wanting to look at the device.17:09
paul_boddieNeither of those should cause a problem, and you can fix them by (1) changing the endpoint descriptor, and (2) editing libmtp's udev rules.17:10
paul_boddiePeople also seem to have problems with modem-manager - it wants anything that declares itself as CDC ACM - but that's not an issue here: the advantage of using Debian instead of Ubuntu. ;-)17:11
paul_boddieSolving the issue with a firmware update is unfortunately my responsibility here.17:12
larscmaybe give cypress a call ;)17:12
paul_boddieThey'd be only too happy to help, I'm sure. :-)17:13
apeleteHi there21:29
apeletelarsc mth: I've been making some progress, the probe function seems to register the glue layer now -> http://paste.debian.net/53739/21:31
apeleteI added a couple of lines to the board file and platform.h to fix the probe -> http://seketeli.fr/git/~apelete/qi-kernel.git/commit/?h=jz4740-udc&id=e7519cf2143a5f1dd525230d2b3cc03e111848d721:33
apeleteI don't know yet why I still get the pagealloc corruption with all the aaaa during boot though21:34
apeletelarsc mth: adn I think I need some help to fix these:21:36
apelete[    1.220000] unable to find transceiver of type USB2 PHY21:36
apelete[    1.230000] musb-hdrc musb-hdrc.0.auto: musb_init_controller failed with status -51721:36
apelete[    1.240000] platform musb-hdrc.0.auto: Driver musb-hdrc requests probe deferral21:36
apeletethe errors above seem to be related to the way musb-hdrc works with the glue layer to "register/create" the device in usbcore, and I don't know much about that21:40
apeletelarsc mth: I would appreciate some help if/when you have time :-)21:41
mthon the jz4770 the "nop" transceiver is used21:57
mthI don't know exactly how that gets selected though21:57
mthI think I saw the nop transceiver in your platform configuration, but is the driver itself also enabled in the kernel config?21:59
apeletemth: I copied the nop transceiver from the jz4770 code, but I dont know its purpose to tell the truth22:20
apeletemth: what is the purpose of the "nop" transceiver, and how is used by the musb driver ?22:21
mthI think it does nothing (nop = no operation), but it is a transceiver, so if the code must have a transceiver then it satisfies that requirement22:26
mth"unable to find transceiver of type USB2 PHY" suggests that a transceiver is mandatory22:26
apeletemth: ok, I see. do you exactly what a USB transceiver is supposed to do in the general case ?22:28
apeletejust asking for my general culture :-)22:28
mthI think it's responsible for putting the actual signals and power on the USB connector22:31
mthit controls the part that is closest to the wires22:32
mthfor the jz4770 we should probably implement a transceiver; currently there the glue code controls the signal levels and power22:33
mthafaik the jz4740 doesn't have software control over that22:34
apeleteso I won't ever need a transceiver in jz4740 other than the nop transceiver ?22:36
apelete<mth> for the jz4770 we should probably implement a transceiver; currently there the glue code controls the signal levels and power22:40
apeleteso that's why there are functions like jz_musb_init_regs() and jz_musb_set_vbus() in the jz4770 code I guess...22:40
mthset_vbus is only needed for host mode, which the 4740 doesn't support22:41
apeletein fact I'm wondering how the jz4740 can do without a transceiver (since I'm going to use "nop" transceiver) AND without such functions that handle "actual signals and power on the USB connector"22:43
mththe register flags that the 4770 glue code sets for transceiver-like functionality simply don't exist on the 474022:43
mththere is a lot less to configure on the 474022:44
apeleteok. I still don't get the whole picture but I'll see how it turns out in the code anyway, then we'll talk again I guess :-)22:46
apeletemth: thanks a lot for helping22:46
--- Wed Oct 9 201300:00

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