**** BEGIN LOGGING AT Sun Oct 12 03:14:13 2008 Oct 12 03:14:23 (that's our logger) Oct 12 03:14:24 uh oh Oct 12 03:14:34 ok first point Oct 12 03:14:36 BobOfDoom: your zImage has done bad things Oct 12 03:14:43 eh? Oct 12 03:14:51 ie - it did nothing Oct 12 03:14:59 no linux console comes up Oct 12 03:15:00 just froze? Oct 12 03:15:03 yup Oct 12 03:15:05 aargggh Oct 12 03:15:11 ok Oct 12 03:15:12 cocboot just got to loading... Oct 12 03:15:15 lets start with that then Oct 12 03:15:20 hehehe Oct 12 03:15:23 Cocoboot = pile of hacks Oct 12 03:15:28 basically Oct 12 03:15:29 not surprising Oct 12 03:15:40 i twiddled wityh a bit of pal dev a lonmg time ago Oct 12 03:15:42 it asks Palm OS to allocate a chunk of memory to copy kernel into Oct 12 03:15:46 palmos == pile of hacks Oct 12 03:15:47 :) Oct 12 03:15:51 Palm OS memory management == major suck Oct 12 03:16:08 it often refuses to allocate anything over 1 to 1.5mb Oct 12 03:16:31 hmmm Oct 12 03:16:34 mwester: backlog sent Oct 12 03:16:39 which is usually the cause for whaty you're seeing (unless the zImage is dead) Oct 12 03:16:40 odd Oct 12 03:16:42 thanks :) Oct 12 03:16:54 the simage i was using before was 1.5m with the initrd built in Oct 12 03:16:56 and its been reliable Oct 12 03:17:02 yours is 1.2 Oct 12 03:17:17 hmm Oct 12 03:17:21 that's pretty odd then Oct 12 03:17:39 * rwhitby can now go to the beach with kids, since logging is now on :-) Oct 12 03:17:41 * BobOfDoom getting worried now Oct 12 03:18:00 let me try again Oct 12 03:18:01 raster: http://kedar.palmlinux.cz/test/cocoboot-git20080702.prc Oct 12 03:18:08 just in case it's an old cocoboot version Oct 12 03:18:09 thank god palmos is a fast booter Oct 12 03:18:17 aaah might be Oct 12 03:19:03 by the way, the latest source is always at http://git.hackndev.com/ Oct 12 03:19:11 we were previously using SF.net SVN Oct 12 03:19:22 but got data loss (disk corruption) problems Oct 12 03:19:37 mwester, raster: we're hoping to make http://www.handhelds.org/moin/moin.cgi/PalmTreo650 up to date with everything we find out Oct 12 03:19:38 so we decided to move everything to our own hosting, where we have better control over stuff Oct 12 03:20:12 let me try... Oct 12 03:20:27 BobOfDoom: is Petr "P3T3" still active? Oct 12 03:20:28 BobOfDoom: aha! another exiter from sf.net Oct 12 03:20:38 move e dev off sf.net long ago thanks to continual cvs outages Oct 12 03:20:54 raster: we moved nslu2-linux.org off there three years ago for the same reason Oct 12 03:20:54 we did the same thing Oct 12 03:21:00 everything but emails lists is off sf.net Oct 12 03:21:01 rwhitby: I don't think so... I never really talked to him much though, we were basically working independently Oct 12 03:21:36 BobOfDoom: MUCH different kernel Oct 12 03:21:39 tiiinuy fot Oct 12 03:21:41 font Oct 12 03:21:45 ok Oct 12 03:21:48 that boots now Oct 12 03:21:52 with tthe new cocoboot Oct 12 03:21:53 raster: great! :) Oct 12 03:22:00 i dont have an initrd... so thats fine Oct 12 03:22:09 btw Oct 12 03:22:19 that white border on the lcd is annoying as hell Oct 12 03:22:23 any idea on how to nix it? Oct 12 03:22:40 I've tried and had no luck Oct 12 03:22:45 rwhitby: i'll help fix it Oct 12 03:22:50 BobOfDoom: bugger. Oct 12 03:22:55 on the other palms yes Oct 12 03:23:02 you jsut put them in 324x324 resolution Oct 12 03:23:05 got to be some pxa register somewhere Oct 12 03:23:10 but it seems the treo is different Oct 12 03:23:39 oooh Oct 12 03:23:40 ewww Oct 12 03:23:44 seeing fb badness with x now Oct 12 03:24:01 yup Oct 12 03:24:03 hmm Oct 12 03:24:12 fb base address is off... Oct 12 03:24:16 looks by.. 324*2 bytes Oct 12 03:24:19 "_ Oct 12 03:24:20 :) Oct 12 03:24:21 ah Oct 12 03:24:43 ie (x fb begins 324 pixels farther from top-left that it should) Oct 12 03:24:46 worked in my older kernel Oct 12 03:24:52 mwester: I just realised that for once this irc channel is going to be mostly active in the .au timezones :-) You'll be the odd one out. Oct 12 03:25:01 raster: http://git.hackndev.com/?p=ato/linux-2.6-arm;a=blob;f=arch/arm/mach-pxa/palmt650.c;h=053a0f3bfda16eb633d774daa2203109ee02081c;hb=HEAD Oct 12 03:25:13 look at static struct pxafb_mode_info palmt650_lcd_modes[] Oct 12 03:25:22 :) That's what the log is for, I guess! Oct 12 03:25:39 at least backlight comes back on now Oct 12 03:25:42 :) Oct 12 03:27:02 I think the settings in P3T3's image are just these (older 2.6.21 branch): http://git.hackndev.com/?p=linux-hnd;a=blob;f=arch/arm/mach-pxa/palmt650/palmt650.c;h=d03483f4575be6b9a6b71608b71a3b2eb6800c66;hb=HEAD Oct 12 03:27:30 hmm Oct 12 03:27:50 it may be worth double checking this against palmos again Oct 12 03:28:05 cocoboot can dump LCD register (via info menu) Oct 12 03:28:08 hmmm Oct 12 03:28:18 frud Oct 12 03:28:22 usbnet not working Oct 12 03:28:28 u dont build usbnet in? Oct 12 03:28:35 it is builtin Oct 12 03:28:39 hmm Oct 12 03:28:44 should be anyway Oct 12 03:28:52 I think.. Oct 12 03:28:55 [161631.063938] usb 2-2: new full speed USB device using uhci_hcd and address 45 Oct 12 03:28:56 [161631.470708] usb 2-2: device not accepting address 45, error -71 Oct 12 03:28:58 bad palm Oct 12 03:29:00 bad Oct 12 03:29:01 yah Oct 12 03:29:12 I think old cocoboot opened USB on boot Oct 12 03:29:14 this is normal flakeyniess? Oct 12 03:29:16 new one doesn't Oct 12 03:29:18 but you can force it Oct 12 03:29:20 hmm Oct 12 03:29:23 i swear i asked it to Oct 12 03:29:26 there's USB console in the menu Oct 12 03:29:32 oh you did? Oct 12 03:29:33 hmm Oct 12 03:29:37 let me try again Oct 12 03:30:10 hmmm Oct 12 03:30:40 while i'm atr it Oct 12 03:30:46 hceck these lcd regs Oct 12 03:31:41 LCCR0 = 0x06000081 LCCR1 = 0x13070d3f LCCR2 = 0x0768013d LCCR3 = 0x04700005 Oct 12 03:31:45 is what I've got Oct 12 03:31:49 now to look up what htey mean Oct 12 03:32:06 yup Oct 12 03:32:08 same here Oct 12 03:32:09 ok Oct 12 03:32:11 and yes Oct 12 03:32:13 thats the problem Oct 12 03:32:42 btw Oct 12 03:32:57 u can see in the consokle that the text is shifted right by 4 pixels Oct 12 03:33:03 or is that 2 Oct 12 03:33:04 ... Oct 12 03:33:20 should it be right on the edge? Oct 12 03:33:30 i thnik so.... Oct 12 03:33:59 * BobOfDoom consults the PXA270 developer manual Oct 12 03:34:11 x is fdefinitely off by 324 from top-left Oct 12 03:34:16 cool Oct 12 03:34:21 u have a copy Oct 12 03:34:22 :) Oct 12 03:34:58 is it available publicly? Oct 12 03:35:09 it used to be, when the XScales were made by Intel Oct 12 03:35:22 when they sold them to marvell, marvell asked them to take down the docs Oct 12 03:35:28 so I still have a copy from back then Oct 12 03:35:49 aha! Oct 12 03:35:52 worked now Oct 12 03:36:06 maybe i didnt do the usb console last boot Oct 12 03:36:15 http://www.nativesystems.inf.ethz.ch/pub/Main/LecturesSystembau/PXA27x_Processor_Family_Developers_Manual.pdf Oct 12 03:36:30 sweet, thx. Oct 12 03:36:47 * raster wget's Oct 12 03:37:00 ah Oct 12 03:37:02 here we go Oct 12 03:37:03 http://pubs.gumstix.org/documents/PXA%20Documentation/ Oct 12 03:37:24 that's got all the various manuals Oct 12 03:38:27 * mwester just can't imagine why that first URL is so slow downloading... ;) Oct 12 03:39:20 bugger Oct 12 03:39:24 got my module_id wrong Oct 12 03:39:28 BobOfDoom: any idea how I get to be able to edit http://www.handhelds.org/moin/moin.cgi/PalmTreo650Hardware?action=edit ? Oct 12 03:39:35 What do we do for a kernel console on the 650 -- the serial connector pins on the connector, or does cocoboot arrange for something else? Oct 12 03:39:35 (to add that gumstix url) Oct 12 03:40:06 mwester: only sane thing i've foiund is usb ethernet gadget Oct 12 03:40:09 but thats not kernel console Oct 12 03:40:11 rwhitby: hit UserPrefs up top right Oct 12 03:40:13 BUT Oct 12 03:40:21 it does allow for vaguely sane poking around remotely Oct 12 03:40:35 rwhitby: then make up an account and hit "Create Profile" Oct 12 03:40:41 AFAIK all registered users can edit Oct 12 03:40:42 i dont get any usb serial devices appearing Oct 12 03:40:43 BobOfDoom: did that, and have logged in. still no edit Oct 12 03:40:45 just ethernet Oct 12 03:40:49 rwhitby: hmm Oct 12 03:40:53 [162048.967983] usb0: register 'cdc_subset' at usb-0000:00:1a.1-2, Linux Device, 9e:32:8a:fc:b5:8a Oct 12 03:41:17 rwhitby: no "EditText" link and the very bottom of the page? Oct 12 03:41:22 rwhitby: (in the gray box) Oct 12 03:41:24 rwhitby: tnx for log Oct 12 03:41:37 raster: any tty devices? Oct 12 03:41:51 raster: ah.. sorry, I only built in the ethernet driver Oct 12 03:41:55 mwester: none that aren't already there before :) Oct 12 03:42:03 BobOfDoom: aha! Oct 12 03:42:04 :) Oct 12 03:42:10 Ok. I'll cut up a treo cable then, and try. Oct 12 03:42:10 doesnt matter Oct 12 03:42:14 if thernet works i'm happy Oct 12 03:42:21 i just need it to bring up wbu0 by defaultr Oct 12 03:42:27 :) I feel blind without a serial console. Oct 12 03:42:41 mwester: be careful, the normal cables don't have the serial pins connected Oct 12 03:42:51 like the pins are missing from the connector Oct 12 03:43:01 aha - Immutable page (last edited 2008-10-12 02:58:22 by AlexOsborne) Oct 12 03:43:02 That's why I was asking -- i'm going to have to locate a full one. Oct 12 03:43:09 eh... Oct 12 03:43:13 Immutable.. Oct 12 03:43:17 The two I have don't have the pins per the doc I hae. Oct 12 03:43:18 hmm Oct 12 03:43:38 now agaian the joy of.. what is it about om-gta0X vs palmt650 in oe that makes it not ifup usb0 on boot... Oct 12 03:43:39 Immutable? ooh -- needs irresistable force, then. Oct 12 03:44:18 BobOfDoom: can you add that gumstix link to the page while we're sorting out edit perms? Oct 12 03:44:18 module loading is a possiblity, but could also be just that usb0 is not "auto" in the network config file. Oct 12 03:44:31 rwhitby: yeah sure Oct 12 03:44:34 thx Oct 12 03:45:03 * rwhitby bbl Oct 12 03:45:07 mwester: it isnt auto Oct 12 03:45:08 thats the problem Oct 12 03:45:11 but what "adds" the auto Oct 12 03:45:12 :) Oct 12 03:45:15 that.. is the trick Oct 12 03:45:20 YOU do. Oct 12 03:45:25 :) Oct 12 03:45:37 in oe figuring out the magic package/config vlaue/dependency indirect goop whatever that does somethign is... a challenge Oct 12 03:45:43 nooo Oct 12 03:45:49 It's part of the initial setup bb files in OE; I'd have to go greppign to find them. Oct 12 03:45:54 i want my image i build to have ot set up out of the box! Oct 12 03:46:05 yeah Oct 12 03:46:06 i know Oct 12 03:46:08 somewhere... Oct 12 03:46:16 hehe! One minute,then -- I muck about with that often, just not sure off the top of my head. Oct 12 03:46:43 rwhitby: I can't see anything about edit permission with my account, maybe this disabld editing by new accounts cause they were having spam issues Oct 12 03:46:51 s/this/they/ Oct 12 03:46:59 mwester: it'd be handy Oct 12 03:47:00 :) Oct 12 03:47:12 i can tell u the magic that makes x choose pointers or not Oct 12 03:47:15 and splashes.. and..... Oct 12 03:47:25 i can als now tell u what evilness sucks psplash on when i dont want it... Oct 12 03:48:18 aha! Oct 12 03:48:19 netbase Oct 12 03:48:55 how about i add a palmt650 for us eh? Oct 12 03:49:45 Sorry for the wait -- netbase.bb does the dirty work. Oct 12 03:49:55 yeah Oct 12 03:49:56 found it Oct 12 03:50:02 i applied my sledgehammer approach Oct 12 03:50:07 grep -r usb0 . Oct 12 03:50:10 :) Oct 12 03:50:36 theres an om-gta01/02 specific for netbase Oct 12 03:50:41 You can tell netbase is a real mess when every machine/distro has its own copy. Oct 12 03:50:41 but none for palmt650 Oct 12 03:50:46 thsu we get the "default" Oct 12 03:50:49 which sucketh Oct 12 03:50:54 Yeah, create one, eveyrone else has! Oct 12 03:50:58 yeah Oct 12 03:51:02 thats what i'm doing Oct 12 03:51:05 and yes Oct 12 03:51:07 it's a mess. Oct 12 03:51:08 but hey Oct 12 03:51:13 i just want my problem solved Oct 12 03:51:20 (We need to redo that to split them up into "classes" of similarly-equiped devices, IMO) Oct 12 03:51:52 yes Oct 12 03:51:58 i'll let others deal with netbase Oct 12 03:52:03 i'm busy cleaning up the x world Oct 12 03:52:08 that is also a nasty hack Oct 12 03:52:13 right now i'm just bringing it up to snuff Oct 12 03:52:19 then it needs definite cleaning Oct 12 03:52:47 (eg the x splash used to be hacked into kdrive-commin - i remvoed x's splash intoa separate package and if its installed and the splash image is there- it's used) Oct 12 03:52:58 tho as such the problem is.. u need to draw a new splash for every resolution Oct 12 03:53:02 and give it a name Oct 12 03:53:03 that sucks Oct 12 03:53:21 i'm tempted to nuke x's splash.ppm and jut go with black Oct 12 03:53:30 let the wm/x initwhatever do a splash as a client app Oct 12 03:53:46 * mwester _LIKES_ to see the boot process, and would rather see no boot splash at all ;) Oct 12 03:54:19 * raster hates the scrolling text goop and lieks his fading in pulsing progress bar Oct 12 03:54:23 with shiny bits Oct 12 03:54:24 :) Oct 12 03:55:18 BobOfDoom: interesting. i think i got my image from AngstromTreo650-v01.zip Oct 12 03:55:24 oh wait Oct 12 03:55:25 no Oct 12 03:55:56 http://blog.mikeasoft.com/2007/06/30/im-in-ur-treo-linuxin-ur-phonez/ Oct 12 03:55:57 that one Oct 12 03:56:05 http://junk.mikeasoft.com/zImage Oct 12 03:56:10 lcd is correctly set up there Oct 12 03:56:19 so something has changed between those Oct 12 03:56:48 hmm Oct 12 03:57:02 the question is... what Oct 12 03:57:10 that'd narrow it down Oct 12 03:57:51 http://www.palmlinux.cz/p3t3/treo650/t650.patch Oct 12 03:57:52 ah Oct 12 03:57:56 that has some LCD chanes Oct 12 03:58:08 s/chanes/changes/ Oct 12 03:58:26 (sorry just got barrier of entry to screwing with the kernel lcd/fb config setup - busy higher up in userland atm) Oct 12 03:59:44 I should have noticed that earlier.. I hadn't been paying much attention to the screen, been mostly using ssh Oct 12 04:00:08 yeah Oct 12 04:00:09 aha! Oct 12 04:00:13 patch Oct 12 04:00:34 ok Oct 12 04:00:38 some keymapfixes Oct 12 04:00:41 nothign major Oct 12 04:00:46 aha! Oct 12 04:00:48 margin! Oct 12 04:00:59 - .left_margin = 20, Oct 12 04:00:59 - .right_margin = 8, Oct 12 04:00:59 + .left_margin = 24, Oct 12 04:00:59 + .right_margin = 4, Oct 12 04:01:17 i'll bet my left buttock its that Oct 12 04:02:06 if u have time - can u revere that bit of the patch Oct 12 04:02:10 rebuild and i can test? Oct 12 04:02:24 (we need to get your tree into oe for me to do this sanely) Oct 12 04:02:24 ? Oct 12 04:02:34 yeah Oct 12 04:02:36 just a sec Oct 12 04:02:52 wesome Oct 12 04:02:56 now x starts with the right dpi Oct 12 04:03:04 and splash set up properly etc. Oct 12 04:04:36 raster: http://meshy.org/~ato/tmp/zImage Oct 12 04:04:58 excellent Oct 12 04:05:02 usb0 is up by default Oct 12 04:05:02 ok Oct 12 04:05:08 try your zim Oct 12 04:06:23 * raster awaits the rootfs rewrite to sd Oct 12 04:07:49 BobOfDoom: btw - any progress on battery lately? Oct 12 04:08:18 raster: nah, I haven't really had a chance to look at it Oct 12 04:08:22 yes Oct 12 04:08:25 it's different to the other palms Oct 12 04:08:25 copnsole looks good again Oct 12 04:08:31 it's more like the neos Oct 12 04:08:35 has smart battery Oct 12 04:08:54 need to find the gpio it's connected to Oct 12 04:08:54 oooh crap Oct 12 04:09:01 and then figure out the protocol Oct 12 04:09:16 it's documented Oct 12 04:09:17 DS2761 Oct 12 04:09:25 oh rad Oct 12 04:09:34 i was about to give up there Oct 12 04:09:39 at least I think that's what the chip in the battery is, I'm not sure where that info came from Oct 12 04:09:42 if u dont know the proto and dont know what gpio... Oct 12 04:10:30 oooh Oct 12 04:10:31 hmm Oct 12 04:10:33 no Oct 12 04:10:36 not fully fixed Oct 12 04:10:38 IIRC there's even a linux driver for it Oct 12 04:10:44 just need to feed it the right GPIO Oct 12 04:10:51 fb is now alignt correctly on the left Oct 12 04:10:56 but.. tis 1 pixel down from the top Oct 12 04:11:00 ah Oct 12 04:11:18 (if it were not for the annoying white border.. i might never have noticed too!) Oct 12 04:11:18 I'll try changing top margin then Oct 12 04:12:05 ok Oct 12 04:12:13 my keyboard no wories Oct 12 04:12:16 workies Oct 12 04:12:23 oh happiness Oct 12 04:12:28 i just ssh directly in with no setup... Oct 12 04:12:29 joy Oct 12 04:12:32 out of the box Oct 12 04:12:40 :) Oct 12 04:12:43 if u want i can pose my rootfs image Oct 12 04:12:46 it's a lot more polished Oct 12 04:13:09 itsh the exact same rootfs userspace i do for my freerunner Oct 12 04:13:19 nice, yeah that'd be good Oct 12 04:13:25 so it will basically get instant-benefits from dual-development Oct 12 04:13:35 and i will be running on both platforms Oct 12 04:13:44 let me sort out this losss of kbd tho Oct 12 04:14:02 * BobOfDoom currently using debian/armel at the moment so he has to do as little compiling as possible.. but it uses tons of space and needs manual configuring of everything Oct 12 04:14:43 30M openmoko-illume-image-glibc-ipk--20081012-palmt650.rootfs.jffs2 Oct 12 04:14:52 53M openmoko-illume-image-glibc-ipk--20081012-palmt650.rootfs.tar Oct 12 04:14:56 raster: cool, so illume runs ok on there? Oct 12 04:14:58 thats my base os image Oct 12 04:15:01 yup Oct 12 04:15:03 runs fine Oct 12 04:15:14 except when i have to do much IO Oct 12 04:15:18 then things suck Oct 12 04:15:23 the touchscreen seems a bit flakey too Oct 12 04:15:28 i think Oct 12 04:15:39 but its a tiny image Oct 12 04:15:49 i write a 250m ext2 image so its mostly empty Oct 12 04:16:03 but its basic and quickyl reproducable Oct 12 04:16:05 from source Oct 12 04:16:26 + .upper_margin = 7, Oct 12 04:16:27 + .lower_margin = 6, Oct 12 04:16:30 thats why i'm more than hapyp to leave the kernel hackign to you for now as i know i have my userspace/image building in hand and can churn out changes fast Oct 12 04:16:31 (instead of 8, 5) Oct 12 04:16:36 aaah Oct 12 04:16:40 was that another patch? Oct 12 04:16:42 re get http://meshy.org/~ato/tmp/zImage Oct 12 04:16:43 nah Oct 12 04:16:47 I just made that change Oct 12 04:17:05 i'm suspicious of that Oct 12 04:17:12 7 Oct 12 04:17:16 but then... Oct 12 04:17:20 5 is also suspicious Oct 12 04:17:23 indeed Oct 12 04:17:25 it's not even! Oct 12 04:17:51 we should be able to steal the settings from palmos Oct 12 04:18:05 it's just a pain to interpret the registers :p Oct 12 04:18:18 * BobOfDoom has a nother look at them Oct 12 04:18:25 hmm Oct 12 04:18:25 ok Oct 12 04:18:29 x is getting the key events Oct 12 04:18:33 but not passing them on Oct 12 04:18:34 grrr Oct 12 04:19:27 let me get xev.. Oct 12 04:20:22 actually Oct 12 04:20:24 let me check Oct 12 04:20:55 beginning of line wait = 20 pixlock cycles.. so that matches left_margin Oct 12 04:21:26 end of line wait = 8 pixclock cycles so that mathes right_margin Oct 12 04:21:51 oooh Oct 12 04:21:52 much better Oct 12 04:21:57 hmm Oct 12 04:22:13 2.5m/sec read rates now from fs Oct 12 04:22:34 beginning of frame wait = 7 pixclock cycles Oct 12 04:22:37 i was getting < 1m before with loopback rootfs on fat Oct 12 04:22:55 670k/sec writes Oct 12 04:23:01 that's the first (from the left) byte of LCCR2 = 0x0768013d Oct 12 04:23:09 better - was getting 120k on looopback rootfs Oct 12 04:23:20 my conclusion: SCREW LOOPBACK Oct 12 04:23:28 (or find a good fix for it in the kernel) Oct 12 04:23:35 BobOfDoom: ok. maybe 7 is ok Oct 12 04:23:43 eh Oct 12 04:23:58 end of frame wait = 0x68 = 104 ? Oct 12 04:26:57 what page in the manual covers these? Oct 12 04:27:14 7-66 Oct 12 04:27:22 section 7.5.4 Oct 12 04:27:48 PDF page 398 Oct 12 04:28:46 I guess it may be that the end of frame wait doesn't matter too much Oct 12 04:28:52 ok Oct 12 04:28:54 found it Oct 12 04:28:54 I have no idea what the LCD protocol looks like Oct 12 04:29:40 ah Oct 12 04:29:42 yeah Oct 12 04:29:50 aha! Oct 12 04:29:51 fiuxed Oct 12 04:29:58 bob! you're a genius Oct 12 04:29:59 it just waits for that many pixclock cycles before sending the next VSYNC Oct 12 04:30:03 raster: I am? Oct 12 04:30:03 fb is back to normal now Oct 12 04:30:07 yay! Oct 12 04:31:25 :) Oct 12 04:32:38 grrr Oct 12 04:32:40 ok Oct 12 04:32:53 x has decided now to cease doing anythign with keyboard/console events Oct 12 04:35:03 hmm Oct 12 04:35:10 must be something in your new kernel images Oct 12 04:35:17 but console definitely emits something Oct 12 04:35:24 read(0, "\34", 256) = 1 Oct 12 04:35:36 hmm Oct 12 04:35:37 read(0, "\234", 256) = 1 Oct 12 04:35:56 thats a definitey key press/release combo if i ever saw them Oct 12 04:35:58 so x gets them Oct 12 04:36:02 then... throws them out Oct 12 04:36:48 IIRC it was working with Debian's X Oct 12 04:36:50 lemme check Oct 12 04:37:01 it was working before too Oct 12 04:37:06 though keymap was foobared Oct 12 04:37:25 ooooh Oct 12 04:37:26 um Oct 12 04:37:27 hmm Oct 12 04:38:18 ii wonder if its vt1... Oct 12 04:39:04 hmm Oct 12 04:39:06 indeed Oct 12 04:39:10 not working for me either Oct 12 04:39:31 hmmm Oct 12 04:39:34 keyboard is /dev/input/event0 Oct 12 04:39:39 x opens it Oct 12 04:39:44 it gets the input Oct 12 04:39:49 strace shows me that Oct 12 04:39:53 but.. then.. does nothing Oct 12 04:40:20 nothing from xev Oct 12 04:40:27 yup Oct 12 04:40:32 but strace x itself Oct 12 04:40:39 u'll see its reading from fd # 0 Oct 12 04:40:56 yeah I see Oct 12 04:40:58 hmm Oct 12 04:42:07 yup Oct 12 04:42:10 works now Oct 12 04:42:12 now... Oct 12 04:42:17 why is it so Oct 12 04:42:38 what'd you do to fix it? Oct 12 04:42:47 or you're trying the other kernel? Oct 12 04:42:57 ok Oct 12 04:43:01 i was telling it to use vt1 Oct 12 04:43:04 thats why Oct 12 04:43:16 instead of the next available Oct 12 04:43:21 and this seemed to screw up keyboard Oct 12 04:43:21 odd Oct 12 04:43:55 BUT Oct 12 04:43:58 it means its an easy fix Oct 12 04:44:02 keymap still screwed Oct 12 04:44:03 :) Oct 12 04:44:19 which keys? Oct 12 04:44:29 is it screwed in the console or just X? Oct 12 04:44:50 numbers Oct 12 04:44:58 the "blue button" bottom-left Oct 12 04:45:05 oh Oct 12 04:45:08 isnt recognised as a "numlock" thing Oct 12 04:45:13 yeah Oct 12 04:45:25 I think it's mapped to CTRL Oct 12 04:45:32 this can be done with some xmods and remaps Oct 12 04:45:37 hmm Oct 12 04:45:37 no Oct 12 04:45:41 right-shift is ctrl Oct 12 04:45:42 i've found Oct 12 04:45:59 right-shift =right ctrl Oct 12 04:46:04 blue thingy = left ctrl Oct 12 04:46:06 IIRC Oct 12 04:46:17 oooh Oct 12 04:46:19 yest it is Oct 12 04:46:23 hmmm Oct 12 04:46:28 i wonder if thats good Oct 12 04:46:38 maybe it should be meta Oct 12 04:46:41 or hyper Oct 12 04:46:50 as u want it to not work like a ctrl Oct 12 04:46:58 but as a different modifier Oct 12 04:47:06 thats ok Oct 12 04:47:11 can do this all in x land Oct 12 04:47:11 :) Oct 12 04:47:29 either way Oct 12 04:47:31 yeah.. I figured CTRL is most useful for the console Oct 12 04:47:38 u want the keyboard to work almsot the same as in palmos Oct 12 04:47:38 but you can remap in the console too Oct 12 04:47:43 yeah Oct 12 04:47:47 but need the numbers Oct 12 04:47:55 and some symbols like = dont exist on the kbd Oct 12 04:48:23 I think I had a mapping file for the console that rayban (one of early devs working on this) did Oct 12 04:48:28 * BobOfDoom tries to find it Oct 12 04:48:50 ok... Oct 12 04:48:56 so how do u want your fs images? Oct 12 04:49:42 tarball? Oct 12 04:49:56 if that's what you mean Oct 12 04:51:02 well ext2 or tarball Oct 12 04:51:06 i was going to ask Oct 12 04:51:07 or jffs2? Oct 12 04:51:12 (if u really want fun) Oct 12 04:51:17 btew Oct 12 04:51:19 btw Oct 12 04:51:34 i guess there is zero hope of getting the disk-on-chip to work Oct 12 04:51:37 that msystems thing is common Oct 12 04:51:44 and they seem to be non-nice in terms of providing info Oct 12 04:51:57 yeah Oct 12 04:52:10 some of the HTC projects reverse engineered it enough to make a read-only driver Oct 12 04:52:24 but they didn't want to try to make a write one without some way to recover Oct 12 04:52:43 hahahaha Oct 12 04:52:49 they dont have devices to "throw away" Oct 12 04:52:50 :) Oct 12 04:52:59 its only 32m Oct 12 04:53:10 to make that useful we'd need to have the bootloader hacked Oct 12 04:53:18 ie possibly a uboot port Oct 12 04:53:23 yep Oct 12 04:53:26 to be honest - i'm happy with the cd card option Oct 12 04:53:29 err Oct 12 04:53:30 sd Oct 12 04:53:33 it works Oct 12 04:53:37 and is dead-easy to reflash Oct 12 04:53:37 :) Oct 12 04:53:45 yep, safe and simple Oct 12 04:56:16 yeah Oct 12 04:56:23 uploading ext2 and tar images now Oct 12 04:56:37 http://handhelds.org/cgi-bin/cvsweb.cgi/linux/kernel26/arch/arm/mach-pxa/htcuniversal/docg3.c?rev=1.1&content-type=text/x-cvsweb-markup Oct 12 04:56:44 ^ there's their driver Oct 12 04:56:54 cool, thanks. :) Oct 12 04:58:44 i'll also upload my sd-card base dir Oct 12 04:59:01 ie with cocoboot, zImage, cocboot.conf and linux.boot.conf Oct 12 04:59:09 just as a quick-start-kit Oct 12 04:59:15 i should documented it Oct 12 04:59:53 but its the "get sd card - make 2 partitions (p1 minimum 4m, p2 minimum 256m) Oct 12 04:59:58 make p1 fat16 Oct 12 05:00:06 untar sdcard-base.tar.gz in it Oct 12 05:00:13 make p2 ext2 (linue) Oct 12 05:00:29 zcat ext2 image onto the partition device Oct 12 05:00:37 take sdcard - put in treo Oct 12 05:00:41 happy booting Oct 12 05:01:30 I guess this it? http://download.enlightenment.org/misc/Illume/Treo-650/2008-10-12/ Oct 12 05:01:37 HAHAa Oct 12 05:01:38 u guessed Oct 12 05:01:41 i just finished upload Oct 12 05:01:45 writign a README Oct 12 05:08:09 ok Oct 12 05:08:10 got a readme Oct 12 05:13:03 BobOfDoom: i notice the ts driver i think has issues Oct 12 05:13:07 i press Oct 12 05:13:15 but the x,y coords of the press are the PREVIOUS location Oct 12 05:13:21 so it registers a press before a move Oct 12 05:13:24 i think... Oct 12 05:13:42 hmm Oct 12 05:14:48 oooh Oct 12 05:14:54 chgeck this out Oct 12 05:15:17 oh no Oct 12 05:15:49 yes Oct 12 05:16:03 MotionNotify event, serial 23, synthetic NO, window 0xc00001, Oct 12 05:16:03 root 0x3b, subw 0x0, time 4054475057, (170,130), root:(170,173), Oct 12 05:16:03 state 0x0, is_hint 0, same_screen YES Oct 12 05:16:03 ButtonPress event, serial 23, synthetic NO, window 0xc00001, Oct 12 05:16:03 root 0x3b, subw 0x0, time 4054475057, (170,130), root:(170,173), Oct 12 05:16:04 state 0x0, button 1, same_screen YES Oct 12 05:16:06 MotionNotify event, serial 23, synthetic NO, window 0xc00001, Oct 12 05:16:08 root 0x3b, subw 0x0, time 4054475075, (204,48), root:(204,91), Oct 12 05:16:10 state 0x100, is_hint 0, same_screen YES Oct 12 05:16:12 MotionNotify event, serial 23, synthetic NO, window 0xc00001, Oct 12 05:16:12 ah Oct 12 05:16:14 root 0x3b, subw 0x0, time 4054475094, (203,48), root:(203,91), Oct 12 05:16:16 state 0x100, is_hint 0, same_screen YES Oct 12 05:16:18 ButtonRelease event, serial 23, synthetic NO, window 0xc00001, Oct 12 05:16:20 root 0x3b, subw 0x0, time 4054475175, (203,48), root:(203,91), Oct 12 05:16:22 state 0x100, button 1, same_screen YES Oct 12 05:16:24 then... Oct 12 05:16:32 hmm Oct 12 05:16:33 no Oct 12 05:16:36 the next isnt interesting Oct 12 05:16:37 but Oct 12 05:16:47 that was a snippet of me tapping just once Oct 12 05:16:52 top-right-ish Oct 12 05:17:01 it got a motion notify for where the mouse was before Oct 12 05:17:07 yeah Oct 12 05:17:07 and a press where it was before Oct 12 05:17:13 then moved to where it really is Oct 12 05:17:15 then the release Oct 12 05:17:22 so the release is at the right spot Oct 12 05:17:40 i would suspect kernel driver at this stage Oct 12 05:17:53 as this is the first time i have seen such a thing across several platforms via tslib Oct 12 05:18:08 yeah, I think that's likely too Oct 12 05:18:19 x is just dumb - it just does whatever tslib reports Oct 12 05:18:23 and tslib isnt that smart either Oct 12 05:18:24 :) Oct 12 05:19:04 i dont have too many things on that image there to test with Oct 12 05:19:06 it's probably cause it gets the pen down event as an interrupt Oct 12 05:19:08 but u'll see the same in debian Oct 12 05:19:14 it must be reporting immediately Oct 12 05:19:20 before it gets the new location Oct 12 05:19:34 no es bueno Oct 12 05:19:39 if u get the down interrupt Oct 12 05:19:50 should know the correct location before reporting Oct 12 05:19:51 :) Oct 12 05:21:22 makes the ts a little... less useful Oct 12 05:21:50 indeed Oct 12 05:22:07 * BobOfDoom has a look at the driver Oct 12 05:26:30 http://git.hackndev.com/?p=ato/linux-2.6-arm;a=blob;f=drivers/input/touchscreen/palmt650-wm97xx.c;h=0040de82b4155cdb8f7388132ccae3400e3b26e9;hb=HEAD Oct 12 05:26:36 wm97xx_acc_pen_down() Oct 12 05:27:10 MODR == AC97 modem FIFO (the chip abuses it to send the touchscreen data efficiently) Oct 12 05:27:58 so it looks like the chip is actually giving us the interrupt before the change in coordinates Oct 12 05:28:05 maybe just skipping the first sample would do Oct 12 05:36:17 sorry Oct 12 05:36:20 was getting my pants on Oct 12 05:36:38 WTF! Oct 12 05:36:45 ts via ac97 modem! Oct 12 05:36:47 eeeevil! Oct 12 05:36:58 haha Oct 12 05:37:19 you can run in a polling mode as well Oct 12 05:37:27 rather than streaming via ac97 modem Oct 12 05:37:49 the continuous mode is supposed to be lower CPU usage though Oct 12 05:38:08 thats fine Oct 12 05:38:17 but... you need to defer Oct 12 05:38:36 it is deferred Oct 12 05:38:48 that's not the irq handler Oct 12 05:38:48 ummm Oct 12 05:38:52 doesnt look like it Oct 12 05:38:56 see wm9712.c Oct 12 05:39:02 and wm97xx-core.c Oct 12 05:39:04 as well Oct 12 05:39:08 that's just the machine specific part Oct 12 05:40:11 hmm Oct 12 05:40:36 * raster reads Oct 12 05:40:36 ooh Oct 12 05:40:39 fancy bootsplash Oct 12 05:40:51 * BobOfDoom trying raster's rootfs Oct 12 05:40:54 BobOfDoom: :) Oct 12 05:40:58 yeah Oct 12 05:41:04 its my universal bootsplash Oct 12 05:41:08 ie identical config/theme Oct 12 05:41:14 multipel devices/resolutions Oct 12 05:41:18 no need to specially re-do it Oct 12 05:41:21 it auto-adapts Oct 12 05:41:33 nice Oct 12 05:42:19 actually the screen on the plam has a nasty gamma Oct 12 05:42:26 the dark greys are black Oct 12 05:42:31 and the light greys are washedout Oct 12 05:42:39 wow... Oct 12 05:42:46 it looks amazing Oct 12 05:42:55 Oh no, the boots ? Oct 12 05:42:56 looks bettre on the freerunner Oct 12 05:43:03 mwester: of course not! Oct 12 05:43:14 let me do a video of it on my treo Oct 12 05:43:33 I hadn't tried your FR image either Oct 12 05:44:32 i needd to uplaod todays build Oct 12 05:44:34 identical to treo Oct 12 05:48:37 ehehe, I love the flip-down time widgets in the alarm Oct 12 05:48:56 :) Oct 12 05:49:08 the joy of abstracting everything into a theme Oct 12 05:49:11 the theme does that Oct 12 05:49:25 code just talks to the theme blob abstractly and says "set time to 12:00:37" Oct 12 05:49:30 theme figures out how to display Oct 12 05:49:32 and animate Oct 12 05:49:33 etc. Oct 12 05:50:10 theme is edje right? Oct 12 05:50:46 I played with it a little bit, it's a neat way of doing things Oct 12 05:51:52 anyway enough drooling, I'll see if I can fix that TS driver Oct 12 05:52:06 BobOfDoom: yup Oct 12 05:52:09 everything is edje! Oct 12 05:52:21 the world... is edje! Oct 12 05:52:26 or will be... soon! Oct 12 05:52:28 muhahaha Oct 12 05:52:32 :D Oct 12 05:53:55 ok Oct 12 05:54:05 lets convert these mjpeg jobs into something useful Oct 12 06:27:44 http://www.rasterman.com/files/illume-freerunner.avi Oct 12 06:27:49 http://www.rasterman.com/files/illume-treo-650.avi Oct 12 06:39:40 raster: wow.. from the video that TS problem is really bad Oct 12 06:39:48 mine only does it occasionally Oct 12 06:39:58 mine does it all the time Oct 12 06:40:06 thats why i noticed Oct 12 06:40:30 i went "wtf... how can that bug be in ecore or evas.. what could i have done that did that? Oct 12 06:40:33 and i mulled Oct 12 06:40:44 and decided that that was hyper-unlikely Oct 12 06:40:49 so i whipepd out xev Oct 12 07:15:00 BobOfDoom: awake? Oct 12 07:18:27 raster: yep Oct 12 07:18:32 * BobOfDoom screwed up Oct 12 07:18:41 the wiki was just plain wrong Oct 12 07:18:44 was about to aks - it seesm the treo is 312mhz when booted Oct 12 07:18:47 and mem bus is 100mhz Oct 12 07:18:49 correct? Oct 12 07:19:17 not sure Oct 12 07:19:21 is that what linux says? Oct 12 07:19:30 well bogomips is 311 Oct 12 07:19:39 and from memory thats indicating 312Mhz Oct 12 07:19:59 yep Oct 12 07:20:15 http://en.wikipedia.org/wiki/Treo_650 Oct 12 07:20:16 yup Oct 12 07:20:17 dmesg Oct 12 07:20:20 Palm OS version 5.4 / 312 MHz Intel PXA270 processor Oct 12 07:20:26 Run Mode clock: 104.00MHz (*8) Oct 12 07:20:26 Turbo Mode clock: 312.00MHz (*3.0, active) Oct 12 07:20:26 Memory clock: 104.00MHz (/1) Oct 12 07:20:26 System bus clock: 104.00MHz Oct 12 07:21:20 PXA doesnt have any video accel worht talking of Oct 12 07:21:23 last i looked... Oct 12 07:21:26 i think it has planes Oct 12 07:21:31 like an overlay plane Oct 12 07:22:07 yup Oct 12 07:22:08 overlay Oct 12 07:22:30 yeah Oct 12 07:22:34 ie Oct 12 07:22:37 mostly useless Oct 12 07:34:31 ah ha! Oct 12 07:35:40 ? Oct 12 07:35:51 (busy writing up some bloggage about all of this) Oct 12 07:36:40 I think I might have fixed it Oct 12 07:36:47 turns out I was wrong Oct 12 07:36:57 unlike all the other palms it doesn't have an interrupt on the treo Oct 12 07:37:36 turns out I mistyped something 2 years ago and mistook something else for the touchscreen interrupt XD Oct 12 07:37:50 so I've just pulled out my broken code Oct 12 07:38:03 and switched it back to straight polling, without the AC97 modem crap Oct 12 07:38:14 and I can no longer reproduce the problem Oct 12 07:38:28 I'll upload a zImage for you to try in a sec, since you were able to produce the problem easier Oct 12 07:39:25 oh radsicles! Oct 12 07:39:46 raster: http://meshy.org/~ato/tmp/zImage Oct 12 07:40:09 I've also made the driver settings appear in /sys/module/wm97xx_ts/paameters Oct 12 07:40:19 s/paa/par/ Oct 12 07:40:23 bleh Oct 12 07:40:28 you get the idea Oct 12 07:40:57 ja ja Oct 12 07:40:58 might be able to tweak, as it seems to be fairly jittery Oct 12 07:41:38 kaafree: hello! :) Oct 12 07:42:34 trying it out... Oct 12 07:46:25 BobOfDoom: much better! Oct 12 07:46:27 all working nicely Oct 12 07:47:22 raster: great! :) Oct 12 07:52:49 http://www.rasterman.com/ Oct 12 07:52:52 there blogged about it Oct 12 07:53:01 btw Oct 12 07:53:07 that svn link has an oe soruce setup u might like Oct 12 07:53:13 as it is what i use to buidl that illume image Oct 12 07:53:14 etc. Oct 12 07:53:15 anyway Oct 12 07:53:17 have to run Oct 12 07:53:18 bbl Oct 12 07:53:50 cya Oct 12 09:39:56 raster: once I get downloads.freesmartphone.org autobuilder building your images, then you can enter into a free agreement with me for hosting of all the sources (which will be done automatically by the autobuilder) Oct 12 10:34:48 raster: I've created a "stable" 2.6.27 branch by merging my development branch into mainline 2.6.27 (I was working off some random pre 2.6.27 git version). I'm testing building in OE at the moment, if it works I'll send you the OE patch so you can build the kernel for your images off that, so you don't have to rely on my binaries. Oct 12 10:35:12 http://git.hackndev.com/?p=linux-2.6;a=summary Oct 12 10:53:03 I'm trying out raster's README, but when I put the dual-partition 2GB SD card into the Treo, it insists on formatting it back to a single partition Oct 12 10:53:45 I'm doing a 1GB FAT16 partition and a 1GB ext2 partition. It's *not* an SDHC card. Oct 12 10:54:14 rwhitby: yeah Palm OS is fussy about the partition Oct 12 10:54:17 BobOfDoom: do you have OE write access? If not, I can commit patches for you until we get it for you. Oct 12 10:54:30 rwhitby: nope I don't have write access Oct 12 10:54:37 rwhitby: try vfat instead Oct 12 10:54:44 or.. Oct 12 10:54:52 hang on lemme check what I'm using Oct 12 10:54:52 ok, we'll work on getting you OE access if you would like it Oct 12 10:55:16 I'm formatting the card in linux, using fdisk in Ubuntu in a VirtualBox session on a laptop running windows :-) Oct 12 10:55:28 rwhitby: cool :) Oct 12 10:55:35 what are you setting the partition id to in fdisk? Oct 12 10:55:40 maybe 1GB is too big for the fat partition or sometjing Oct 12 10:55:41 I mean the partition type Oct 12 10:55:43 6 for Oct 12 10:55:55 6 for partition type Oct 12 10:56:05 83 for the second one Oct 12 10:56:37 Disk /dev/sdb: 2032 MB, 2032664576 bytes Oct 12 10:56:38 41 heads, 40 sectors/track, 2420 cylinders Oct 12 10:56:38 Units = cylinders of 1640 * 512 = 839680 bytes Oct 12 10:56:38 Disk identifier: 0x00000000 Oct 12 10:56:38 Device Boot Start End Blocks Id System Oct 12 10:56:39 /dev/sdb1 1 1210 992180 6 FAT16 Oct 12 10:56:41 /dev/sdb2 1211 2420 992200 83 Linux Oct 12 10:56:53 Device Boot Start End Blocks Id System Oct 12 10:56:53 /dev/sde1 1 5 9734 6 FAT16 Oct 12 10:56:54 /dev/sde2 6 255 488250 83 Linux Oct 12 10:56:55 then mkdosfs on /dev/sdb1 Oct 12 10:56:59 hmm Oct 12 10:57:58 Device Boot Start End Blocks Id System Oct 12 10:57:58 /dev/sde1 1 120 98380 6 FAT16 Oct 12 10:57:58 /dev/sde2 121 2000 1541600 83 Linux Oct 12 10:58:00 /dev/sde3 2001 2420 344400 82 Linux swap / Solaris Oct 12 10:58:03 that's my other card Oct 12 10:58:08 I use mkfs.vfat Oct 12 10:58:10 for that one Oct 12 10:58:17 try that maybe Oct 12 10:59:42 nope, same result :-( Oct 12 10:59:59 dang Oct 12 11:00:17 wonder if 2GB is too big Oct 12 11:00:42 nope, looks like you're using a 2GB card above Oct 12 11:00:43 I'm using 2GB cards Oct 12 11:00:44 yeah Oct 12 11:00:55 hmm Oct 12 11:01:27 well.. another last resort thing to try would be let it format it and then resize the partition with parted or something Oct 12 11:01:50 trying partitioning in a real linux box (a freecom fsg3 router running debian) Oct 12 11:14:11 BobOfDoom: still alive? Oct 12 11:14:34 BobOfDoom: gparted did the trick Oct 12 11:14:43 rwhitby: great :) Oct 12 11:14:47 Sleep_Walker: yep Oct 12 11:14:56 I'm doing some GPIO hunting on my Treo680, but I have no access to hh.org wiki Oct 12 11:14:59 great :) Oct 12 11:15:19 Sleep_Walker: ah, you too? damn.. they must have closed it off Oct 12 11:15:36 It was edited by you - you can't modify it? Oct 12 11:15:50 Sleep_Walker: I can.. but no new accounts can seemingly Oct 12 11:16:05 rwhitby: let me know when you have things all building fine Oct 12 11:16:06 I've fixed backlight GPIO, Power Detect, Silent Switch, SD Detect Oct 12 11:16:19 (by "you too", I meant rwhitby ran into the same problem) Oct 12 11:16:24 the problem is... several sources are not upsteram in oe well several bb's and patches Oct 12 11:16:28 but I can't find WM9712 IRQ... how? Oct 12 11:16:28 Sleep_Walker: if you like I can edit it for you Oct 12 11:16:38 ah! Oct 12 11:16:40 hehe Oct 12 11:16:44 I jsut had that problem Oct 12 11:16:50 there is no WM9712 IRQ Oct 12 11:16:52 I was mistaken Oct 12 11:16:54 BobOfDoom: yay! patches! Oct 12 11:16:58 play sound and find what is changing and wasn't? Oct 12 11:17:01 I spent ages this afternoon trying to figure it out Oct 12 11:17:07 :) Oct 12 11:17:09 rwhitby: btyw - i had that too.l i let it format it once Oct 12 11:17:15 you've got Treo680 too? Oct 12 11:17:15 turns out palmos just polls the touchscreen Oct 12 11:17:15 then repartitioned again and it didnt complain Oct 12 11:17:22 no, on the 650 Oct 12 11:17:26 I see Oct 12 11:17:30 but I assume it's the same on 680 Oct 12 11:17:34 raster: still just getting the binaries running, then will work on reproducing from source in OE proper. Oct 12 11:17:39 well probably Oct 12 11:17:40 (including cocoboot source in OE) Oct 12 11:17:52 it's optional anyway, it just reduces CPU usage of touchscreen Oct 12 11:18:11 BobOfDoom: aha! also just polling. well.. i guess what works works... Oct 12 11:18:32 BobOfDoom: btw - the 680 has more ram - right? Oct 12 11:18:40 nah Oct 12 11:18:43 it's got more flash I think Oct 12 11:18:44 no? Oct 12 11:18:47 no Oct 12 11:18:52 palmos doesn't need RAM Oct 12 11:18:59 it can only run one program at once Oct 12 11:19:08 true Oct 12 11:19:08 so there's no point in putting extra ram in a palm os device ;) Oct 12 11:19:13 well actually Oct 12 11:19:15 not quite true Oct 12 11:19:20 Chatter downloads emails while I'm doing other stuff ... Oct 12 11:19:23 the simple answer - y4es Oct 12 11:19:39 yes Oct 12 11:19:43 the complex answer... you can have a background task of some sort via some copoperative shimming into some os bits Oct 12 11:19:44 it'd use for more cache Oct 12 11:19:48 BobOfDoom: more ram :) Oct 12 11:19:50 ah Oct 12 11:19:52 yeah Oct 12 11:19:53 64 MB Oct 12 11:19:54 i dont remember the exact details - but i did read up about it a bit Oct 12 11:19:57 Sleep_Walker: oh really? Oct 12 11:20:01 Sleep_Walker: awesome Oct 12 11:20:05 I didn't realise that Oct 12 11:20:05 well, Linux reports that Oct 12 11:20:35 slp thats what i thought Oct 12 11:20:39 # Oct 12 11:20:39 # 32 MB onboard SDRAM split between system memory, dynamic memory and cache Oct 12 11:20:46 according to wikipedia Oct 12 11:20:55 linux can be wrong Oct 12 11:20:55 ah Oct 12 11:20:57 i didnt bother about a 680 as i didnt see any reports that said linux is working anywhere near as well as on the 650 Oct 12 11:20:58 that explains it! Oct 12 11:21:01 your boot problems! Oct 12 11:21:02 so i'd rather go the 650 for now Oct 12 11:21:03 either way Oct 12 11:21:06 32m is ok for me Oct 12 11:21:11 e is nice and lean Oct 12 11:21:12 :) Oct 12 11:21:19 BobOfDoom: fine :D Oct 12 11:21:25 Linux is told how much RAM by cocoboot Oct 12 11:21:37 Cocoboot makes a completly dodgie guess Oct 12 11:21:43 based on what Palm OS says Oct 12 11:21:45 GPIO changes commited to git.hackndev.com/srv/git/linux-hnd Oct 12 11:21:55 cause palmos reports weird numbers Oct 12 11:22:01 I know Oct 12 11:23:02 Sleep_Walker: adding mem=32m might make your booting more stable Oct 12 11:23:34 to command line Oct 12 11:23:43 and that also explains that leds and keyboard backlight weird problem on boot :b Oct 12 11:23:50 hmm Oct 12 11:23:54 speaking of kbd backlight Oct 12 11:24:07 whats that connected to? Oct 12 11:24:12 ASIC6 Oct 12 11:24:17 (and i hope not the shin bone) Oct 12 11:24:29 undocumented HTC chip of doom Oct 12 11:24:30 and that is connected to... /sys/....？ Oct 12 11:24:31 :) Oct 12 11:24:35 umm Oct 12 11:24:41 HAHAHA Oct 12 11:24:48 aha! the htc chip of doom Oct 12 11:24:49 that one Oct 12 11:24:53 I don't think I've ported the driver to 2.6.27 yet Oct 12 11:24:56 * BobOfDoom checks Oct 12 11:25:35 no I haven't Oct 12 11:25:37 but I'll do it right now Oct 12 11:25:40 that can also fix ASIC6 problems in Linux :) Oct 12 11:25:40 should be easy Oct 12 11:25:59 * Sleep_Walker is booting Oct 12 11:26:32 Sleep_Walker: if that helps we can hardcode it into cocoboot so it doesn't use the detected values and you don't have to stick that on the command-line Oct 12 11:26:38 BobOfDoom: AHHAHAHA no hurry! Oct 12 11:26:43 i was just wondering Oct 12 11:27:00 BobOfDoom: I first try if it fixes my problems Oct 12 11:27:06 as its always on... Oct 12 11:27:33 raster: hmm.. yeah, guess we should turn it off when we blank the LCD Oct 12 11:28:09 BobOfDoom: could you please have a look here? http://git.hackndev.com/?p=linux-hnd;a=blob;f=arch/arm/mach-pxa/palmt680/palmt680.c;h=90db311b01e0363d5ae20000f5c76070250afde7;hb=2259c410bdf2811f5107685800921b3143fd7b64#l127 Oct 12 11:28:26 I've just hacked to solve backlight problem Oct 12 11:28:39 that line 130 Oct 12 11:28:53 BobOfDoom: yeah. though i'd leave it to userspace - just have a /sys/... thing for it Oct 12 11:29:11 raster: booting your image now Oct 12 11:29:25 Sleep_Walker: eh? intensity is inverted? Oct 12 11:29:25 in fact.... Oct 12 11:29:31 BobOfDoom: yup :( Oct 12 11:29:36 q - whats the plan for power saving? Oct 12 11:29:46 rwhitby: yay! boot! Oct 12 11:29:59 maybe I can just fidle some other settings... Oct 12 11:30:09 raster: ah.. that's tricky.. because a lot of the power management stuff is tied up in that aforementioned chip of doom.. Oct 12 11:30:19 yeah Oct 12 11:30:25 just trying a bit of suspend Oct 12 11:30:29 ah Oct 12 11:30:30 seems the systsme suspended Oct 12 11:30:36 levt me with a blinking kbd backlight Oct 12 11:30:41 but no resume... :( Oct 12 11:30:46 yeah Oct 12 11:31:01 tho personaklly i'd liek to skip suspend/resume Oct 12 11:31:07 I'm about to release a new version of cocoboot to fix that Oct 12 11:31:12 how good a shape is the pxa freq clocking? Oct 12 11:31:20 ie linux support Oct 12 11:31:24 eg go down to 0 clock Oct 12 11:31:26 or like 1 mhz Oct 12 11:31:27 ... Oct 12 11:31:56 and then do all the powersaving by explicitly powering down devices Oct 12 11:31:57 raster: not sure... I haven't tried it in 2.6.27... it sucked in 2.6.21 (it messed up the LCD) Oct 12 11:32:03 eg bt, backlight/screen etc. etc. Oct 12 11:32:08 hmm Oct 12 11:32:09 but that would be really nice if we can do it that way Oct 12 11:32:11 * rwhitby types on the linux command line on the Treo650 for the very first time Oct 12 11:32:12 lcd is synced to it Oct 12 11:32:16 hm, with that mem=32m it behaves _very_ slowly Oct 12 11:32:27 raster: yeah.. we patched it.. it's probably been fixed in mainline by now as well Oct 12 11:32:28 i'll take a messed lcd... if we do a cafeat Oct 12 11:32:44 you just need to scale the LCD clock at the same time Oct 12 11:32:44 "lcd is messed on ultra-low-clock-mode. u want lcd working. u need 312mhz" Oct 12 11:33:11 Sleep_Walker: what userspace are you using? Oct 12 11:33:19 well that's weird Oct 12 11:33:22 rwhitby: usbnet? Oct 12 11:33:23 OPIE from auryn Oct 12 11:33:24 maybe it does have 64mb RAM then Oct 12 11:33:38 Sleep_Walker: do you have swap? Oct 12 11:33:53 raster: nope, forgot to click the option. Terminal keyboard. Oct 12 11:33:55 hmmm Oct 12 11:33:55 nope :D Oct 12 11:34:00 i have 32m ram on my treo Oct 12 11:34:01 Sleep_Walker: maybe boot it in 64mb, and try to allocate and write to say 50mb of memory Oct 12 11:34:02 it wasn't needed before Oct 12 11:34:02 and no swap Oct 12 11:34:05 and its doing ok Oct 12 11:34:14 rwhitby: aaaah Oct 12 11:34:22 I have to check syntax for that parameter Oct 12 11:34:23 rwhitby: in cocoboot rememebr to do that Oct 12 11:34:24 then boot Oct 12 11:34:29 then unplug usb/data cable Oct 12 11:34:31 plugin Oct 12 11:34:32 and bingo Oct 12 11:34:37 u'lkl get a ethernet usb dev Oct 12 11:34:48 ssh in exactly as with gta01/02 Oct 12 11:34:52 maybe big M? :b Oct 12 11:35:21 Sleep_Walker: are u using the ext2 file image on fta16 Oct 12 11:35:24 or a partition? Oct 12 11:35:28 raster: gotta work out how to get that routed into the virtualbox session ... Oct 12 11:35:39 I've done it with the neo, so shouldn't be too hard Oct 12 11:35:50 raster: squashfs-lzma boot as live linux Oct 12 11:35:56 (from fat) Oct 12 11:36:07 rwhitby: virtualbox session? Oct 12 11:36:07 ooh Oct 12 11:36:09 no wonder Oct 12 11:36:12 you're using ramdisk? Oct 12 11:36:19 yup Oct 12 11:36:27 and no swap Oct 12 11:36:28 Sleep_Walker: hmm that'll be a problem. i found the ext2 image at least on sd-card had HORRID io Oct 12 11:36:33 but it was working like shame Oct 12 11:36:44 i moved to ext2 on a 2nd sd partition and my read rates doubled Oct 12 11:36:47 well I'm surprised Oct 12 11:36:49 raster: it's squashfs, not ext2 Oct 12 11:36:49 my write rates went up 4x Oct 12 11:36:53 well if it works, keep it at 64 then Oct 12 11:36:56 i know Oct 12 11:37:03 I guess it should have kernel paniced if it was 32mb Oct 12 11:37:05 and it worked before great Oct 12 11:37:11 raster: my laptop runs windows, and I run ubuntu in a virtualbox session. all my other linux machines are embedded. Oct 12 11:37:13 but u're paying a price for it being a loopback file on fat Oct 12 11:37:33 raster: he's not loopback, he's unpacking into ramdisk Oct 12 11:37:35 raster: yup, not so much Oct 12 11:37:39 BobOfDoom: nope Oct 12 11:37:42 BobOfDoom: ooooh! Oct 12 11:37:43 just mounting Oct 12 11:37:43 no? Oct 12 11:37:44 ramisk Oct 12 11:37:46 oh Oct 12 11:37:48 aufs Oct 12 11:37:56 ooh Oct 12 11:37:56 SLAX way Oct 12 11:37:56 ok Oct 12 11:38:00 I'm confused Oct 12 11:38:04 how does that work? Oct 12 11:38:15 is it like a ramdisk but loads on demand Oct 12 11:38:16 check kEdAR's init scripts if interested Oct 12 11:38:17 ? Oct 12 11:38:28 ramdisk for writing Oct 12 11:38:31 rwhitby: aaaah! so not a native linux laptop - thus it isnt getting all usb devices routed to it. gotcha Oct 12 11:38:39 squashfs-lzma modules for reading Oct 12 11:38:50 unified with aufs (another union fs) Oct 12 11:38:55 Sleep_Walker: ah gotcha Oct 12 11:39:04 interesting setup XD Oct 12 11:39:06 pretty fast Oct 12 11:39:13 raster: I'll probably just plug it into the embedded router running debian instead Oct 12 11:39:44 It died before I started cat /proc/meminfo :b Oct 12 11:39:57 rwhitby: whataever works for you :) Oct 12 11:40:10 maybe I should use some memmap parameter Oct 12 11:40:20 rwhitby: the pxa27x usb gadget driver has a note on it saying it doesn't work with RNDIS (or whatever the windows usb ethernet thing is called) Oct 12 11:40:28 * raster doesnt like his valuable ram being filled with ramfs's Oct 12 11:40:33 root@palmt650:~# Oct 12 11:40:39 rwhitby: cause the hardware is apparantly buggy and can only export one device and once Oct 12 11:40:58 BobOfDoom: yeah, I never try and connect things like this into windows itself Oct 12 11:41:06 raster: then think about memory upgrade :b Oct 12 11:41:08 the router works :-) Oct 12 11:41:23 Sleep_Walker: kind of hard to given its a treo650... Oct 12 11:41:29 not like i'll be soldering on any ram Oct 12 11:41:30 :) Oct 12 11:41:48 either way - i like the challeng of making sure e+everything runs fine in 32m Oct 12 11:41:57 my first linux box was a p120 with 32m Oct 12 11:42:06 takes me back to some nostalgia... Oct 12 11:42:06 :) Oct 12 11:44:00 BobOfDoom: memorymap.txt should show me all memory addresses used, right? Oct 12 11:44:15 raster: when you click the up arrows on the alarm clock, the leaves show move upwards rather than downwards ;-) Oct 12 11:44:23 s/show/should/ Oct 12 11:44:28 Sleep_Walker: it shows you the main mappings, so yeah, you could use that to tell I guess Oct 12 11:44:54 (it won't show some of the small pages as it only parses large mapping entry) Oct 12 11:45:03 raster: make that when you hit the down arrows (to decrease the number) Oct 12 11:45:19 BobOfDoom: it should show chip select with attached memory module, or I am wrong? Oct 12 11:45:31 Sleep_Walker: it should Oct 12 11:45:36 oh wait? Oct 12 11:45:40 you mean for SDRAM? Oct 12 11:45:47 yup Oct 12 11:45:58 should be 0xa0000000 in the right hand column Oct 12 11:46:02 it is Oct 12 11:46:05 where does that go up to? Oct 12 11:46:11 but it should show size Oct 12 11:46:16 a130 0000 Oct 12 11:46:18 and then Oct 12 11:46:27 a3e0 0000 Oct 12 11:46:30 up to Oct 12 11:46:35 ah. Oct 12 11:46:35 a3f0 0000 Oct 12 11:46:39 yeah that's 64mb then Oct 12 11:46:43 fine Oct 12 11:46:49 lucky bastard :) Oct 12 11:46:54 but with hole in the middle, right? Oct 12 11:47:26 like I said it doesn't show small paegs Oct 12 11:47:32 maybe it allocates heap with small pages or something Oct 12 11:47:38 on 650 Oct 12 11:47:38 it show Oct 12 11:47:59 BobOfDoom: what's the status on bluetooth support? Oct 12 11:48:07 a0000000-a0d00000, a0f00000-a1900000, a1a00000-a1e00000 Oct 12 11:48:26 rwhitby: they are meant to do that Oct 12 11:48:31 (flip down always) Oct 12 11:48:37 thats just how i designed the graphics Oct 12 11:48:41 and the animation Oct 12 11:48:45 its not done in code Oct 12 11:48:53 raster: yeah, I'm commenting on the design, not the implementation ;-) Oct 12 11:49:06 rwhitby: I've found some of the GPIOs and I think I know how to talk to it Oct 12 11:49:13 rwhitby: but I haven't put a driver together and tested it yet Oct 12 11:49:17 also thats quite a meaty bit of theme "code" Oct 12 11:49:18 real clocks with those flip-over leaves actually flip the leaves up when you set the time backwards Oct 12 11:49:22 (ie in the .edc for edje) Oct 12 11:49:29 rwhitby: it's connected via ASIC6 (chip of doom)( Oct 12 11:49:29 its good enough for swisscom Oct 12 11:49:34 i have lots of other fish to fry Oct 12 11:49:39 in the temrinal run: Oct 12 11:49:42 elementary_test Oct 12 11:49:46 raster: if you're going to go overboard with bling, you should at least get it right ;-) Oct 12 11:49:50 thats the test suite so far for elementary Oct 12 11:50:01 rwhitby: but it's a standard UART I think, so we can just point linux's UART driver at the base address, and it might work Oct 12 11:50:08 oh not going overboard with bling... yet! Oct 12 11:50:12 so should I tell memmap=32M@a0000000 memmap=32M@a3e00000 Oct 12 11:50:45 Sleep_Walker: nah Oct 12 11:50:50 BobOfDoom: I have bluetooth access points in my home and work specifically for the treo, so BT support is a priority for me :-) Oct 12 11:50:50 Sleep_Walker: I don't think it is holey Oct 12 11:50:57 BobOfDoom: i suspect there is some magic that means there is major renewed interest in linux on treo... Oct 12 11:50:57 Sleep_Walker: I think it's just missing the mappings Oct 12 11:50:59 of late Oct 12 11:51:04 all good for linux and treos Oct 12 11:51:05 :) Oct 12 11:51:10 BobOfDoom: I see Oct 12 11:51:22 Sleep_Walker: a3e00000 to a3f00000 isn't 32mb Oct 12 11:51:23 so memmap=64M@a0000000 Oct 12 11:51:28 that reminds me Oct 12 11:51:39 i need to brew up a keymap for x that makes the keyboard work proeprly Oct 12 11:51:43 ie numbers and symbols Oct 12 11:51:43 :) Oct 12 11:51:44 BobOfDoom: I know, but it is important where it is Oct 12 11:52:05 oh geez - I forgot to even try the real keyboard in the Terminal! Doh! Oct 12 11:52:10 raster: I need the same for console :( Oct 12 11:52:14 raster: 32mb == 0x02000000 if I'm not being stupid Oct 12 11:52:15 err Oct 12 11:52:19 Sleep_Walker: I mean Oct 12 11:52:26 BobOfDoom: right Oct 12 11:52:45 raster: picture me poking on the pop-up virtual keyboard on a treo, when there is a perfectly good hardware keyboard on the device. what an idiot! Oct 12 11:52:54 Sleep_Walker: so if it starts at 0xa0000000 then 64mb should go up to a3ffffff Oct 12 11:53:09 right Oct 12 11:53:23 which matches that a3e00000 Oct 12 11:53:25 it's near the end Oct 12 11:53:29 right Oct 12 11:53:36 so memmap=64M@a0000000 Oct 12 11:53:45 raster: magic? Oct 12 11:54:18 raster: heh.. it's openmoko's fault :p making people start thinking about linux phones again Oct 12 11:56:54 raster: for me it is caused by having a bit more time, being overread, overanimed and disgusted by T3 :b Oct 12 11:59:20 heh, there ALREADY IS mem=32M in kernel command line Oct 12 11:59:29 it was added as default in .config Oct 12 11:59:46 Sleep_Walker: won't cocoboot overwrite .config? Oct 12 11:59:49 or does it add to it? Oct 12 12:00:08 I think it adds Oct 12 12:00:13 let me check it Oct 12 12:00:18 I've got dmesg somewhere Oct 12 12:00:41 ok, overwrited Oct 12 12:05:27 hm, with mem=64M it doesn't even get to FrameBuffer Oct 12 12:05:55 O.o Oct 12 12:06:00 I'm totally confused then Oct 12 12:06:10 unless palmos is just mapping it for no good reason Oct 12 12:06:14 that'd be like palmos :p Oct 12 12:07:24 well, I'd rather ask - should there be also 32MB of flash? :b Oct 12 12:07:31 (connected same way) Oct 12 12:08:38 64 MB onboard NOR Strataflash (part of the PXA CPU) Oct 12 12:08:43 64 MB onboard NAND storage as user-available stored non-volatile memory Oct 12 12:08:53 maybe Oct 12 12:09:02 :O Oct 12 12:09:02 it's not a DoC like the 650 Oct 12 12:09:06 128? Oct 12 12:09:13 not 128? Oct 12 12:09:21 maybe wikipedia is just full of it Oct 12 12:09:28 we need a better source :p Oct 12 12:10:13 this agrees: http://handheld.softpedia.com/devices/palmOne/palmOne-Treo-680--Lowrider-398.shtml Oct 12 12:11:07 if it's not DoC, does that mean it would be easier to access the internal flash? Oct 12 12:11:14 rwhitby: yes Oct 12 12:11:18 I haven't tried to fullfill it yet since I've got 4gb sdhc :) Oct 12 12:11:31 rwhitby: should be able to use standard linux flash drivers Oct 12 12:11:55 Sleep_Walker: damnit, you're making me want one... that's sound like quite a nice device Oct 12 12:12:15 Sleep_Walker: next we'll find out it's got wifi.. :P Oct 12 12:12:15 great :D Oct 12 12:12:30 it has on board :b Oct 12 12:12:52 but palmos doesn't use it Oct 12 12:13:02 it's like T3's led red :b Oct 12 12:13:10 haha Oct 12 12:13:43 btw. it should have bcm2045 Oct 12 12:13:48 so bluetooth 2.0 Oct 12 12:15:12 and palmos probably doesn't use that Oct 12 12:16:04 amfg Oct 12 12:16:23 hard reseted device, ran cocoboot and it cause another reset Oct 12 12:17:06 * rwhitby puts in a $150 bid for a treo 680 on ebay.au Oct 12 12:18:12 rwhitby: heh.. I was just looking there too Oct 12 12:18:23 hmm - $152.50 and 3 days to go :-( Oct 12 12:18:23 :D Oct 12 12:18:54 if everyone will have Treo680, then kernel will be ready very soon Oct 12 12:19:50 http://stores.ebay.com.au/RIX-SMARTPHONES-and-RIX-HOBBIES has quite a few Oct 12 12:21:16 * rwhitby now has 2 bids in, but expects to get neither Oct 12 12:21:40 BobOfDoom: btw - PALMT650_ASIC6_VIRT and PALMT650_ASIC6_SIZE - how got you this? Oct 12 12:22:06 Sleep_Walker: made them up Oct 12 12:22:22 virtual address to map to in linux.. just chose something that wasn't being used Oct 12 12:22:32 and btw. "pxa2xx_ac97_reset: cold reset timeout (GSR=0x0)" got this in dmesg Oct 12 12:22:35 ok Oct 12 12:22:40 so this has no need to change Oct 12 12:22:51 yep Oct 12 12:24:04 not sure what that message is about, maybe it's just the ac97 chip complaining the nice palm os driver it was talking to mysteriously got assasinated and a dodgy hippy linux driver appeared in its place Oct 12 12:24:43 well, I don't know, but sound is not working yet :b Oct 12 12:26:47 Sleep_Walker: alsamixer? Oct 12 12:27:04 I'm overalsamixered Oct 12 12:27:22 hmm Oct 12 12:27:27 all on and up, switching switches Oct 12 12:27:36 mplayer seems to play Oct 12 12:27:42 there's a GPIO to turn on power to the external speaker on 6500 Oct 12 12:27:43 err Oct 12 12:27:45 650* Oct 12 12:27:54 hmmm, ok Oct 12 12:27:57 but you can at least hear through the earpiece and headset Oct 12 12:28:12 I haven't treied earphones yet Oct 12 12:28:49 I lent one PDA masta hacka from work reduction between 2.5 and 3.5 jack :) Oct 12 12:29:26 since I've got BT headphones - there was no problem Oct 12 12:32:12 * BobOfDoom tries building python with -mthumb Oct 12 12:33:08 gotta be able to cut frameworkd down so its usable without swap Oct 12 12:33:24 * Sleep_Walker hopes that PalmOS is clever enough to power down external speaker on plugging headphones Oct 12 12:41:49 night guys Oct 12 12:41:56 rwhitby: night Oct 12 12:43:56 rwhitby: gn Oct 12 12:57:20 hm, I've got incredibly slow mahcine now :( Oct 12 12:57:25 I have to figure why Oct 12 12:59:47 it was easy :b Oct 12 12:59:52 memory management fault :b Oct 12 13:01:08 BobOfDoom: linux kernel knows it's position and size in memory, right? Oct 12 13:04:12 Sleep_Walker: the position and size of the kernel code itself you mean? yeah I imagine it would :p Oct 12 13:04:35 I dunno where to look it up though if thats what you mean Oct 12 13:06:15 hm.. alignmnet trap - it is related to memory, isn't it? Oct 12 13:06:27 rwhitby: hahahah! yeah.... the proboem is keymap Oct 12 13:06:36 even then.. missing things like = for example Oct 12 13:06:42 with mem=64M it has much less errors :b Oct 12 13:06:46 still possibl;y need some kind of extended virtual input Oct 12 13:06:53 Sleep_Walker: sounds like it Oct 12 13:08:26 BobOfDoom: i've always had a deep interest in linux and embedded. and for me at leats phoens are almost the ultimate form of embedded. always on, always with you, always connected to a network, and small. they are the ubiqutous device... and imho linux will never "succeed" on the desktop. that means take over ms's rule on the desktop workhorses for your average house and business. Oct 12 13:08:41 but man will it be able to rule on the device... like the phone :) Oct 12 13:08:52 raster: I compiled in the cpu frequency scaling, it's saying scaling_min_freq == 104 Mhz Oct 12 13:09:05 so I guess the really low freq modes aren't supported, at least not like that Oct 12 13:09:08 min 104? Oct 12 13:09:10 bugger Oct 12 13:09:10 ok Oct 12 13:09:12 thats useless Oct 12 13:09:24 i know omap has a "0 clock" mode Oct 12 13:09:31 registers go into self-refresh Oct 12 13:09:34 and ram does too Oct 12 13:10:26 ah Oct 12 13:10:30 according to the manual Oct 12 13:10:45 there's "idle mode": allows stopping CPU clock during period of inactivity Oct 12 13:10:45 let's see Oct 12 13:10:49 * raster is alro reading manual Oct 12 13:10:52 256k of sram Oct 12 13:10:54 hmm Oct 12 13:11:04 BobOfDoom: aha! thats what we want Oct 12 13:11:07 idle mode Oct 12 13:11:22 iterrupts shoudl still wake it up back to full clock Oct 12 13:11:32 imho thats probably the most common mode of operation Oct 12 13:11:35 yep Oct 12 13:11:37 stop-clock in idle Oct 12 13:11:48 so that plus tickless kernel should be pretty low power Oct 12 13:11:52 (and just before shut down all peripheral devices u dont need or low-power-mode them) Oct 12 13:12:10 then any interrupt thousl wake up in a few ms at most Oct 12 13:12:15 like < 100 Oct 12 13:12:18 got doc Oct 12 13:12:27 figure out what needs dealing with Oct 12 13:12:28 deal with it Oct 12 13:12:31 go straight back to idle Oct 12 13:12:39 if its a "user facing" thing Oct 12 13:12:48 just delay going into idle for a bit Oct 12 13:12:59 as they are likely interacting and want max performance Oct 12 13:13:12 only thign i knwo of that wouldnt want this is thngs like mp3 playback Oct 12 13:13:16 screen off - no ui Oct 12 13:13:25 but respond to buttons and compute to pdecode/play Oct 12 13:13:49 (well ok if u have a gps and other interestign hadrware maybe stats gathering on gps locations and wifi hotspots...) Oct 12 13:13:52 anyway Oct 12 13:13:56 idle mode sounds good Oct 12 13:14:04 if we can power ram down int a low power mode too Oct 12 13:14:09 and all other perihperals Oct 12 13:14:31 hmm nothing about idle mode in arch/arm/mach-pxa, so I gues it's not implemented in Linux Oct 12 13:14:58 but it can't be that hard can it? :) Oct 12 13:14:59 well... theres the challenge Oct 12 13:15:00 :) Oct 12 13:15:20 u know what the 4x64k of sram is used for? Oct 12 13:15:28 seems interesting as its not a cache Oct 12 13:15:32 dedicated sram Oct 12 13:15:46 you mean the on-CPU RAM? Oct 12 13:15:49 yup Oct 12 13:15:50 at the moment linux doesn't know about it Oct 12 13:15:56 palmos uses it for early bootloader Oct 12 13:16:01 but nothing after that AFAIK Oct 12 13:16:07 know where its mapped? Oct 12 13:16:32 see page 1150 Oct 12 13:16:44 0x5c000000 Oct 12 13:17:14 woot Oct 12 13:17:25 0x4800_0004 Oct 12 13:17:32 thats the baby u want for idle clock for main sdram Oct 12 13:18:20 was thinking Oct 12 13:18:29 sram should have much lower latency than external sdram Oct 12 13:18:45 would make for a grat scratch-buffer for large-data-bloc compute tasks Oct 12 13:18:49 ie software-managed cache Oct 12 13:19:00 yeah Oct 12 13:19:20 in fact for mp3 playeback this kind fo stuff would be good Oct 12 13:19:23 I guess it could be exposed so userspace can mmap it or something Oct 12 13:19:44 wake up - ready some mp3 data from fs layer into sram Oct 12 13:19:52 set sdram to self-reres Oct 12 13:19:59 decode inside sram and pumep out - sti at 104mhz Oct 12 13:20:04 yeah Oct 12 13:20:11 well i was just thining mmap /dev/mem Oct 12 13:20:15 but thats just anti-social Oct 12 13:20:19 ie - what x does Oct 12 13:20:20 :) Oct 12 13:21:09 :D Oct 12 13:21:26 but yes Oct 12 13:21:41 i suppose a quick and dirty kernel device to allow single open of it and mmap of it Oct 12 13:21:47 ie only 1 process at once can open Oct 12 13:21:51 covers all 4 banks Oct 12 13:22:12 allow non-roto access by default and on open memset to 0 Oct 12 13:22:18 so any new process always gets 0 ram Oct 12 13:22:32 u knwo what sucks Oct 12 13:22:36 intel got out of the arm market Oct 12 13:22:52 yeah :( Oct 12 13:22:52 their arms were pretty decent Oct 12 13:22:59 (in the scheme of things) Oct 12 13:23:10 ah... there's also a 13Mhz mode Oct 12 13:23:20 oooh now that sounds yummy Oct 12 13:23:54 ok Oct 12 13:23:57 which needs special LCD handling, but it says the LCD will still work at 26mhz Oct 12 13:24:01 0x48000004 Oct 12 13:24:07 I mean the picxel clock is 26mhz Oct 12 13:24:12 page 299 Oct 12 13:24:24 that baby is what we want for shunting memory down to low power mode Oct 12 13:24:46 as such if we are in 13mhz mode Oct 12 13:24:47 frankly Oct 12 13:24:50 i expect lcd is off Oct 12 13:24:55 because backlight sucks power Oct 12 13:25:02 and if we are squeezing arm to 13mhz Oct 12 13:25:05 but keeping backlight on Oct 12 13:25:07 we're stupid Oct 12 13:25:08 :) Oct 12 13:25:23 admittedly if we dynclock on the fly Oct 12 13:25:26 true Oct 12 13:25:36 and this i a mode we drop into veyr oftne with no glitches going in/out Oct 12 13:25:37 pity the screen isn't very readable with backlight off Oct 12 13:25:53 (i expect there to be glithces clokcing up and down ad we reset lcd clocks and other clocks on the system to match) Oct 12 13:26:07 so i would assume we change clock not so often Oct 12 13:26:40 ie we clock down to 13mhz when screen is blanked (that means lcd out clock off so lcd loses contents and backlight off - some ioctl() to /dev/fb handles that - forgot the details) Oct 12 13:27:05 oooh Oct 12 13:27:13 bit 29 lets u 1/4 clock ram Oct 12 13:27:18 ah that's cool it can auto power down the sdram when part of its not being used Oct 12 13:27:22 ie i guess 26mhz Oct 12 13:28:08 bit 30 -> 1 i thnik is low power self-refresh Oct 12 13:28:18 ok, it seems that alignment problem is there with or without mem=64M Oct 12 13:28:50 must set mdrefr to 0 Oct 12 13:29:20 Entering Alternate Bus Master Mode Refresh Control ? Oct 12 13:29:29 bit 31 seems to be similar Oct 12 13:29:29 hmm Oct 12 13:29:30 confused Oct 12 13:29:35 "The PXA27x processor supports an alternate master on the SDRAM memory bus. " Oct 12 13:29:44 well sdram refrsh "not" performed Oct 12 13:29:52 until counter timed out Oct 12 13:30:30 that sounds like youve got two chips talking to same RAM and transitioning control from one to the other Oct 12 13:30:48 oh wait Oct 12 13:30:49 no Oct 12 13:30:51 bit22 Oct 12 13:30:53 thats the one Oct 12 13:30:55 sorry Oct 12 13:30:56 my bad Oct 12 13:30:58 * Sleep_Walker has playing music into headphones :) Oct 12 13:31:01 yes Oct 12 13:31:04 check bit 22 Oct 12 13:31:04 that makes sense Oct 12 13:31:05 yeah Oct 12 13:31:10 heeh Oct 12 13:31:16 eg u have a gfx chip (like the most loved glamo) and pxa Oct 12 13:31:33 so clock to 13mhz Oct 12 13:31:37 or idle no-clokc Oct 12 13:31:41 BobOfDoom: do you know the way how to route it to external speaker? Oct 12 13:31:42 and just before set bit22 Oct 12 13:31:49 (and just after clear 22) Oct 12 13:31:52 (ie coming out) Oct 12 13:32:12 Sleep_Walker: yeah, sort of Oct 12 13:32:43 Sleep_Walker: see this pic: http://bp3.blogger.com/_7zXusCFNoZI/Rq88xZBZuEI/AAAAAAAAAAc/eDExqrvF6QQ/s1600-h/pamt650-audio-wiring.png Oct 12 13:32:55 :D Oct 12 13:33:25 Sleep_Walker: except I got the MICs backwards Oct 12 13:33:34 Sleep_Walker: some more notes here: http://hhtinker.blogspot.com/2008/10/sucess-two-way-phone-call-with-treo-650.html Oct 12 13:34:43 raster: anyway, we'll probably have to figure out how to power down the peripherals first or it's probably not worth it ;) Oct 12 13:36:15 I've got somewhere screwed USB I think Oct 12 13:36:33 lsusb on PC side show nothing :( Oct 12 13:36:48 BobOfDoom: sure. like kbd backlight Oct 12 13:36:49 gsm Oct 12 13:36:50 bt Oct 12 13:37:02 almost everything else is on the pxa anyway Oct 12 13:37:08 audio, lcd, .... Oct 12 13:37:09 true Oct 12 13:37:20 gsm, kbd are covered Oct 12 13:37:27 then theres bt Oct 12 13:37:35 thats about it... or have i missed something? Oct 12 13:37:45 not even sure we have bt up and working? Oct 12 13:37:47 bt.. well we don't have a driver for it yet so just turn off in palmos Oct 12 13:37:52 this is where we just parallelise Oct 12 13:38:00 get cpu/ram/core clockign right Oct 12 13:38:07 let each peripheral be worked on on its own Oct 12 13:38:18 yep Oct 12 13:38:23 heh, got same GPIO like Tre650 :) Oct 12 13:38:24 good point Oct 12 13:38:24 and as devices get full power control- bring them into the fold and manage them Oct 12 13:38:38 we DO need some infra to do that tho Oct 12 13:38:38 Sleep_Walker: 116? Oct 12 13:38:44 114 - for USB Oct 12 13:38:49 ie a shared userspace daemon where u can request a powerstate... Oct 12 13:38:57 fso is kind of tryng to solve that Oct 12 13:39:04 not sure its doing it the best way... Oct 12 13:39:05 but hey Oct 12 13:39:22 mm.. having python daemons in our 32mb of ram :( Oct 12 13:39:32 I start zhone and frameworkd and I have to go into swap Oct 12 13:40:12 BobOfDoom: i'm not happy with that idea Oct 12 13:40:24 BobOfDoom: i do have a c daemon that handles power stuff Oct 12 13:40:26 ompower Oct 12 13:40:45 wrote it before fso did anything Oct 12 13:40:50 it does work Oct 12 13:40:53 but its very simple Oct 12 13:40:57 can recycle it... Oct 12 13:41:21 cool Oct 12 13:41:24 it allows for multiple system resources Oct 12 13:41:29 and state requests Oct 12 13:41:37 ie client X asks for "cpu on" Oct 12 13:41:42 client b asks for "cpu off" Oct 12 13:41:45 cpu on wins Oct 12 13:41:50 once A removes his cpu on Oct 12 13:41:54 if no other cpu ons are there Oct 12 13:41:56 cpu if "off" Oct 12 13:42:12 in ompower tho backlight/screen are tied to "cpu" Oct 12 13:42:21 so they kind of move together Oct 12 13:43:27 at least thats infra to build on. Oct 12 13:43:44 yeah Oct 12 13:44:05 ah ha! my bitbake is finally up to the kernel Oct 12 13:45:12 it'll take a while longer Oct 12 13:45:12 :) Oct 12 13:45:25 true Oct 12 13:45:32 it's gonna take ages just to checkout Oct 12 13:45:39 since it's pulling it from the net Oct 12 13:45:45 fresh git clone Oct 12 13:45:50 o sh*t Oct 12 13:46:10 Sleep_Walker: don't tell me you broke it? Oct 12 13:46:12 it was maybe muted now with mut Oct 12 13:46:15 mute Oct 12 13:47:55 there is so much switches! :E Oct 12 13:48:01 raster: btw, new zimage with leds, keypad backlight http://meshy.org/~ato/tmp/zImage Oct 12 13:48:14 raster: it's just full-on or full-off though, need to get around to doing the brightness stuff Oct 12 13:48:25 oh it does vibrator too Oct 12 13:48:50 oooh Oct 12 13:49:02 vibrations! Oct 12 13:49:03 oooh oooh Oct 12 13:49:30 Sleep_Walker: just turn everything on, full volume, then twiddle with the multi-select boxes Oct 12 13:49:46 Sleep_Walker: you should be able to find something that works Oct 12 13:50:11 BobOfDoom: external speaker was probably powered down Oct 12 13:50:35 and it had direction out so not setable Oct 12 13:50:40 (116) Oct 12 13:50:48 so I have to find this GPIO first :( Oct 12 13:51:23 we should make some nice tool for doing GPIO stuff Oct 12 13:51:33 palmos programming is so annoying though :( Oct 12 13:52:03 I can try Oct 12 13:54:23 ideally you just want to ignore anything with AF, then record state of everything, let user do something like plug in a cable or whatever, then scan again and show you what changed Oct 12 13:56:52 * Sleep-Walker hates Network Manager Oct 12 13:59:31 raster: bitbakes probably gonna take all night, and it at least seems to parse the bbfile okay and is cloning the right thing Oct 12 13:59:34 raster: http://meshy.org/~ato/tmp/0001-palmt650-Use-the-2.6.27-hackndev-kernel-branch-inst.patch Oct 12 13:59:58 raster: in theory that should get OE to build the exact same zImage I linked to above Oct 12 14:00:23 BobOfDoom: btw Oct 12 14:00:25 kernel booted Oct 12 14:00:27 now Oct 12 14:00:33 where is that kbd lichgt control Oct 12 14:00:38 /sy/class/leds Oct 12 14:00:50 /sys Oct 12 14:00:50 ok Oct 12 14:00:52 saving patch Oct 12 14:00:54 late here Oct 12 14:01:06 oooh Oct 12 14:01:07 palmt650:keypad Oct 12 14:01:08 I think Oct 12 14:01:11 or something like that Oct 12 14:01:13 nice Oct 12 14:01:30 so it's just full-on/off .. we can do brightness, just need to figure out the right number Oct 12 14:01:30 the top led too and sweet sweet sweet vibrations Oct 12 14:01:33 s Oct 12 14:01:47 hmm Oct 12 14:01:53 why palmt650: ? Oct 12 14:02:06 imh i'd love it if... they were more generic Oct 12 14:02:14 hmm Oct 12 14:02:17 ie anything with a vibra has a /sys/class/leds/vibra/... Oct 12 14:02:25 so userspace doesnt have to be re-coded per device Oct 12 14:02:43 yeah, I just followed the convention that the other drivers had but yeah... that makes sense Oct 12 14:02:48 ooh vibra works Oct 12 14:02:49 joy Oct 12 14:02:53 aaah Oct 12 14:02:54 particularly in that you're never going to use the same driver on a different device Oct 12 14:02:54 ok Oct 12 14:02:55 hmm Oct 12 14:03:01 so there's no chance of conflict Oct 12 14:03:03 then i guess its a matter of challenging convetion Oct 12 14:03:14 yeah Oct 12 14:03:24 i would think it'd be better virtualised Oct 12 14:03:32 vibra probably shoudl not be in led's Oct 12 14:03:41 really maybe /sys/class/vibra Oct 12 14:03:57 and leds just remofe palnt650: Oct 12 14:04:07 and then that probably covers most phones/devices Oct 12 14:04:18 led may be better named tho as "power" Oct 12 14:04:23 and "warning" Oct 12 14:04:26 or "charging" Oct 12 14:04:30 mm Oct 12 14:04:31 ie the logical intention of the led Oct 12 14:04:49 ie software decided that blinking the power led to indicate email is fine.. then let it Oct 12 14:05:21 ok, I'm off Oct 12 14:05:23 g'night all Oct 12 14:05:37 gn, BobOfDoom Oct 12 14:07:36 is working IrDA on Treo650? Oct 12 14:07:49 BobOfDoom: night man! great work! Oct 12 14:07:56 hats off to ya Oct 12 14:31:04 heh, 116 is GPIO for headphones detection Oct 12 14:41:09 raster: that GPIO 20 on Treo650 Oct 12 14:41:17 is it LCD power of backlight power? Oct 12 14:41:44 Sleep_Walker: not sure Oct 12 14:41:51 but it is exposed in /sys/... Oct 12 14:42:00 raster: can you tried it? Oct 12 14:42:03 raster: can you try it? Oct 12 14:42:07 /sys/class/backlight Oct 12 14:42:19 raster: it's not what I'm speaking about :b Oct 12 14:42:26 ok Oct 12 14:42:34 maybe it is :b Oct 12 14:42:40 well it does work Oct 12 14:42:55 backlight does vary with breightness Oct 12 14:43:24 it works for me too Oct 12 14:44:08 hm, bob is not touching any GPIO Oct 12 14:44:55 touch touch my gpio! Oct 12 14:44:56 :) Oct 12 14:44:59 hmm Oct 12 14:45:05 looks like we want standby mode... Oct 12 14:46:00 ? Oct 12 14:46:12 power management Oct 12 14:46:15 * Sleep_Walker wants standby mode, but you probably already can do that Oct 12 15:51:36 it seems that IrDA power is also on ASIC6 :( Oct 12 15:51:52 I have to figure out that ASIC6 stuff Oct 12 20:06:19 OMFG Oct 12 20:06:27 wm97xx-ts use polling! :E Oct 12 21:24:49 hey :) Oct 12 21:33:08 hi Oct 12 22:11:12 Sleep_Walker: I've added your GPIOs 0, 15, 38, 113 to http://www.handhelds.org/moin/moin.cgi/PalmTreo680Hardware Oct 12 22:11:27 heh :D Oct 12 22:11:43 I'm filling that table... Oct 12 22:12:11 ahh.. Oct 12 22:12:20 I figured out where wiki permissions are set Oct 12 22:12:22 http://www.handhelds.org/moin/moin.cgi/EditorGroup Oct 12 22:13:21 "Created initial group of people allowed to edit pages. Editable by AdminGroup." -- JameyHicks Oct 12 22:13:24 http://www.handhelds.org/moin/moin.cgi/AdminGroup Oct 12 22:13:36 ah, you can get PaulSokolovsky to add you Oct 12 22:14:03 ok Oct 12 22:16:04 BobOfDoom: http://sleepwalker.hackndev.com/treo-table2.txt :) Oct 12 22:17:25 unfortunately I can't find any IR power (no GPIO difference in PalmOS between on/off) Oct 12 22:18:03 yeah.. I suspect ASIC does most of the power management :( Oct 12 22:18:45 and I still don't know how to work with that :( Oct 12 22:19:07 so can you dump it? Oct 12 22:19:18 hmm Oct 12 22:19:33 well, if it is CS2 as on Treo650 Oct 12 22:19:51 dumping that address cause reset Oct 12 22:19:57 Sleep_Walker: http://trac.hackndev.com/projects/general/wiki/ASIC6 Oct 12 22:19:58 ah Oct 12 22:20:05 I know that page Oct 12 22:20:07 can you post memorymap.txt ? Oct 12 22:20:28 sure Oct 12 22:21:02 http://sleepwalker.hackndev.com/memorymap.txt Oct 12 22:21:24 but there is only one address mapped Oct 12 22:21:39 08000000-08000000 -> 08000000-08000000 Oct 12 22:21:41 :/ Oct 12 22:21:55 hmm Oct 12 22:23:56 well it's mapped so I don't see why dump 08000000 should get you a reset Oct 12 22:24:00 hmm Oct 12 22:24:04 unless it's protected Oct 12 22:24:46 I have no idea :( Oct 12 22:25:02 maybe I can touch and play in linux... Oct 12 22:25:50 but I was quite confused with difference in LEDs and keyboard backlight? Oct 12 22:26:26 ? Oct 12 22:27:14 GPIO20 - On - green led, off - red led (probably previously on) Oct 12 22:27:26 GPIO24 - keyboard backlight Oct 12 22:27:39 it is controled with GPIOs Oct 12 22:27:42 oh.. Oct 12 22:28:08 that's odd Oct 12 22:28:19 does the 680 blink LEDs while in suspend? Oct 12 22:28:36 cause it can't do that without some external chip Oct 12 22:28:48 heh Oct 12 22:28:51 I don't know Oct 12 22:29:03 I'd rather not suspend it Oct 12 22:29:08 nah Oct 12 22:29:08 I mean Oct 12 22:29:10 palmos suspend Oct 12 22:29:15 on 650 Oct 12 22:29:17 if modem is on Oct 12 22:29:18 tada! Oct 12 22:29:24 on your "turn it off" green led blinks Oct 12 22:29:25 slowly Oct 12 22:29:29 to show you its still using power Oct 12 22:29:54 BobOfDoom: does palmos actually "suspend" Oct 12 22:29:59 or more accurately Oct 12 22:30:02 use deep-sleep? Oct 12 22:30:34 BobOfDoom: it seems that it can blink when suspended Oct 12 22:30:41 raster: I think it uses "sleep mode" Oct 12 22:30:56 I'm not 100% sure though Oct 12 22:31:03 hmm Oct 12 22:31:04 (red led - when connecting to GSM network) Oct 12 22:31:08 it definitely does suspend in some way Oct 12 22:31:20 i'm thinking we should give idle or standby modes a go Oct 12 22:31:21 since if you write to 0xa0000000 it won't wakeup properly Oct 12 22:31:27 (that's the resume vector from bootloader) Oct 12 22:31:53 raster: standby should just work.. although we might need to configure wakeup source Oct 12 22:32:14 raster: I set it to wake when the charger is plugged/unplugged but I haven't tested that Oct 12 22:32:36 BobOfDoom: how did you found GSM_WAKE? Oct 12 22:32:39 echo standby > /sys/power/state Oct 12 22:33:32 Sleep_Walker: umm.. the output? hmm.. I think it changed when I turned the phone on and off Oct 12 22:35:46 BobOfDoom: hmm Oct 12 22:35:50 standby kinda works... Oct 12 22:36:02 nothgin wakes it up tho except pwr un/replug Oct 12 22:36:11 on wake usbnet isnt working Oct 12 22:36:12 raster: yeah.. just needs configuring Oct 12 22:36:22 ah that's right Oct 12 22:36:23 device appears Oct 12 22:36:28 need to break usbnet connection on suspend Oct 12 22:36:32 but other end closes Oct 12 22:36:35 otherwise it gets stuck or something Oct 12 22:36:43 yeah Oct 12 22:36:49 cant ssh back in after that either Oct 12 22:37:04 also going into standby is slow Oct 12 22:37:11 well trakes a second or so Oct 12 22:37:12 :( Oct 12 22:37:37 yeah.. that's linux shutting down all the hardware drivers Oct 12 22:37:41 I think Oct 12 22:38:00 yeah Oct 12 22:38:08 thats the biggest problem with the whole "suspend" stuff Oct 12 22:38:19 the kernel having to flush buffers and shut everything down to a known state Oct 12 22:38:23 and on resume do the reverse Oct 12 22:38:31 makes for a very slow "on/off" Oct 12 22:38:39 where u really want it nigh instant Oct 12 22:38:40 yeah Oct 12 22:38:43 eg 100ms or so Oct 12 22:38:48 so whne incoming call happens Oct 12 22:38:52 or bt irq Oct 12 22:38:56 or user presses button Oct 12 22:39:09 things is up and kicking and responding night immediately with "incoming call" Oct 12 22:39:13 or whatever Oct 12 22:39:14 :) Oct 12 22:42:22 suspend-to-mem (using bootloader) does near to instant on/off Oct 12 22:43:36 BobOfDoom, hi, long time no see Oct 12 22:45:30 marex: hi :) Oct 12 22:46:02 marex: ummm.. suspend still takes a while Oct 12 22:46:22 raster, nah ... suspend to mem is pretty fast Oct 12 22:46:31 at least on palmt5 Oct 12 22:46:45 from linux? Oct 12 22:46:50 yea Oct 12 22:48:17 marex: raster wants it instant Oct 12 22:48:35 BobOfDoom, well that's what suspend to mem is ... Oct 12 22:48:49 really? Oct 12 22:48:50 hmm Oct 12 22:48:52 you hit the button and the device goes zZzzzz Oct 12 22:48:57 takes a about 1 s on treo Oct 12 22:49:00 BobOfDoom, for me ... yes Oct 12 22:49:03 :O Oct 12 22:49:35 * raster tests Oct 12 22:50:05 oh dear... Oct 12 22:50:22 no way Oct 12 22:50:24 I resumed and / has disappeared XD Oct 12 22:50:29 its about 6 or so seconds Oct 12 22:50:36 from echo mem > /sys/power/state Oct 12 22:50:36 -bash: /bin/ls: No such file or directory Oct 12 22:50:40 to backlight finally off Oct 12 22:51:04 instant to me means 100ms or less Oct 12 22:51:04 :) Oct 12 23:01:32 fine - GPIOs 2,7,8 or GSM dependant :) Oct 12 23:04:37 * rwhitby welcomes Elleo (Elleo wrote http://blog.mikeasoft.com/2007/07/01/openmoko-on-a-treo-650/) Oct 12 23:05:53 Elleo = Mike? Oct 12 23:06:21 yep Oct 12 23:06:26 great :) Oct 12 23:06:36 his blog was inspirative :b Oct 12 23:07:00 yeah Oct 12 23:07:12 thats where i started off getting my treo up Oct 12 23:07:15 but its a bit old Oct 12 23:07:25 but i have it on track now Oct 12 23:08:49 yeah, Mike told me in an email today that it was over a year ago that he last fiddled with linux on the treo, but he's gonna lurk here anyway. he doesn't have a treo 650 handy at the moment. Oct 12 23:09:22 rwhitby, hi, dont I know you from somewhere ? Oct 12 23:09:37 marex: nslu2-linux, OE, openmoko ? Oct 12 23:10:24 marex: nice work on the t650 upstream push, BTW Oct 12 23:11:51 oh well Oct 12 23:11:56 thosew with handy treos can continue Oct 12 23:12:01 i need to do a proper kemap... Oct 12 23:12:05 that wasnt my doing, I just ported it to .27, BobOfDoom fixed the remaining issues there Oct 12 23:12:22 rwhitby, btw OE is probably the place, yea Oct 12 23:12:27 and then ponder how to solve extended input Oct 12 23:12:36 like for symbols that dont exist on the kbd Oct 12 23:16:08 ie Oct 12 23:16:11 palmos to get = Oct 12 23:16:21 u need to type blue+s (to get -) Oct 12 23:16:28 then press alt to select alternate symbols Oct 12 23:16:31 = is one of them Oct 12 23:18:46 you can do that using some userspace app and uinput Oct 12 23:19:10 gnight Oct 12 23:19:54 gn Oct 12 23:22:29 aaah! Oct 12 23:22:32 found it! Oct 12 23:22:37 keymap Oct 12 23:22:40 just a sec Oct 12 23:23:07 * Sleep_Walker is also interested Oct 12 23:23:19 I want to go to bed and program there :b Oct 12 23:23:29 (on Treo of course) Oct 12 23:24:11 http://meshy.org/~ato/tmp/treo-keymap/ Oct 12 23:24:17 I don't remember how to use it though Oct 12 23:24:28 hmm Oct 12 23:25:17 BobOfDoom: forbidden Oct 12 23:25:18 :) Oct 12 23:25:22 doesnt matter Oct 12 23:25:27 need to do it in x with xmodmap anyway Oct 12 23:25:34 wont help u on console tho Oct 12 23:25:44 but then.. if u are in consolke land i'd expect u to use serial or usbnet Oct 12 23:25:51 ./bin/treokmap:loadkmap < /treo.bmap Oct 12 23:26:56 should be unforbidden now Oct 12 23:27:02 true Oct 12 23:27:26 and still dont have keys for %^_=\|]}[{~`;<> Oct 12 23:27:37 http://meshy.org/~ato/tmp/treo-keymap/Keys.txt Oct 12 23:27:37 thos symbols dont exist on the treo kbd Oct 12 23:27:41 need to emulate them somehow Oct 12 23:28:22 mm Oct 12 23:28:26 some of them are covered Oct 12 23:28:48 although I guess it means you can't to ctrl + [ or whatever Oct 12 23:28:48 hmm Oct 12 23:29:32 not those Oct 12 23:29:38 othjers are Oct 12 23:29:45 poh wait Oct 12 23:29:48 ` is covered Oct 12 23:30:00 of that list Oct 12 23:30:02 but the rest - no Oct 12 23:32:20 how is this map used? Oct 12 23:34:02 I see Oct 12 23:34:04 loadkmap < /treo.bmap Oct 12 23:34:08 I think loadkmap is a busybox applet Oct 12 23:34:25 fine Oct 12 23:34:34 good for console Oct 12 23:34:35 wont help x Oct 12 23:34:48 x should be looking at kbd in raw/semi-raw mode Oct 12 23:48:59 doesn't work for me :/ Oct 12 23:49:04 gn, guys Oct 12 23:49:29 gn **** ENDING LOGGING AT Mon Oct 13 02:59:57 2008