qi-bot | [commit] Xiangfu: nanonote-files: wpan.nn: add option for start dirtpan (master) http://qi-hw.com/p/openwrt-packages/595140e | 02:21 |
---|---|---|
xiangfu | hmm.. the atusb cannot build as module. because the irq_set_chip_and_handler function | 03:41 |
xiangfu | ERROR: "irq_set_chip_and_handler_name" [drivers/ieee802154/spi_atusb.ko] undefined! | 03:41 |
wpwrak | if it must be a module, then you'll have to find a way to hack around this issue | 03:49 |
wpwrak | in the future, with the improved driver, these things will disappear | 03:49 |
wpwrak | so a hack should be fine until then :) | 03:50 |
xiangfu | ok :) | 04:02 |
qi-bot | [commit] Xiangfu: nanonotefiles: make inittab work with minimal config system (master) http://qi-hw.com/p/openwrt-packages/1ec3c83 | 08:44 |
viric | larsc: thank you a lot for your support! I managed to send a patch upstream that 'works for me' | 08:46 |
qi-bot | [commit] Xiangfu: xburst: qi_lb60 select the nanonote slash screen (master) http://qi-hw.com/p/openwrt-xburst/5d69bcb | 08:52 |
qi-bot | [commit] Xiangfu: remove xburst target borken (master) http://qi-hw.com/p/openwrt-xburst/3a25183 | 08:52 |
qi-bot | [commit] Xiangfu: urjtag: fix typo (master) http://qi-hw.com/p/openwrt-packages/f5eefbc | 09:17 |
larsc | viric: nice trick reusing the existing handlers :) | 09:31 |
viric | it took me a while to figure all out :) | 09:34 |
viric | now I've a *segfault* in webkit. | 09:35 |
viric | hm there is a null pointer when there should not be. | 09:39 |
viric | who knows. | 09:39 |
viric | at least webkit runs further. | 09:40 |
lindi- | I have been using chromium on openmoko for quite some time already | 12:03 |
viric | lindi-: is that mips? | 12:59 |
lindi- | no, armel | 13:00 |
lindi- | but you probably have same bugs with unaligned accesses | 13:01 |
viric | well, I was using webkit 1.4.0. Maybe a newer webkit works better. And the troubles were in the unaligned-access-emulation of mips. | 13:02 |
viric | specifically. | 13:02 |
viric | lindi-: and it's related to FPU, that the openmoko maybe does not have | 13:06 |
lindi- | ok | 13:09 |
mth | unaligned access emulation is a horrible thing anyway; it's much better to fix the program doing the access | 13:31 |
viric | of course | 13:32 |
viric | but it becomes a matter of performance, then. Not a matter of "the program does not run" | 13:32 |
viric | and with an 'echo' to a debugfs file, you can make all programs sigbus, if you want, on unaligned access. It becomes up to the kernel user. | 13:33 |
mth | yeah, but I think that trying to make broken programs run is not a good idea, because it decreases the chance of them getting fixed | 13:43 |
mth | also, the resulting system is more complex, like you found out when there was a bug in the unaligned access emulation | 13:44 |
viric | then maybe you'd like the sigbus behaviour to be the default, and the emulation not default | 13:45 |
viric | but it should not be about not having that code in the kernel at all | 13:45 |
viric | Imagine a super-complex package has unaligned accesses, and the producers of the package do not give a penny for your platform. | 13:46 |
viric | (hypothetical case :) | 13:46 |
mth | the emulation can be useful even if only for logging where in the program the unaligned accesses come from | 13:47 |
mth | but indeed I wouldn't want it enabled by default | 13:47 |
mth | same for floating point emulation: I'd rather have the program crash so that I know that softfloat support failed | 13:48 |
Action: DocScrutinizer05 idly wonders why /ns help in this chan and now is the old version still, while /ns help in another chan and 2h ago showed some new details >> | 15:58 | |
DocScrutinizer05 | [16.06.2012 16:43:29] [Notice] -NickServ- If a registered nick is not used by the owner for 150 days, | 15:58 |
DocScrutinizer05 | [16.06.2012 16:43:29] [Notice] -NickServ- NickServ will drop the nickname, allowing it to be reregistered. | 15:58 |
DocScrutinizer05 | 1h ago | 15:58 |
viric | I'm trying fpu code in that loongson2f... and all I compiled looks broken :) | 16:13 |
viric | lame has lots of nans (have to be emulated) | 16:13 |
viric | ffmpeg encoding with libvorbis looks quite mad too; I don't know if due to decoding an mp3, or encoding the ogg. | 16:14 |
viric | I wonder what other floating point code I could run to test | 16:15 |
qi-bot | [commit] Paul Cercueil: ASoC: JZ4740: delay activation of the DAC to work around a sound bug. (jz-3.4) http://qi-hw.com/p/qi-kernel/ef84c71 | 17:34 |
larsc | Ayla: that's an interresting way to implement a msleep() ;) | 17:38 |
Ayla | isn't it? :) | 17:40 |
Ayla | I actually wasn't aware of msleep()... | 17:40 |
Ayla | I know mdelay, not msleep | 17:40 |
larsc | msleep sort of does what you just implemented | 17:41 |
Ayla | ok, I'll edit it then | 17:42 |
qi-bot | [commit] Paul Cercueil: ASoC: JZ4740: Use msleep() to wait before enabling the DAC. (jz-3.4) http://qi-hw.com/p/qi-kernel/5d267d6 | 18:01 |
Ayla | mth: that means we could replace all mdelay() by msleep() on the SLCD panels code | 18:01 |
Ayla | so that the other threads can continue when the panel is initializing | 18:02 |
Ayla | that was a request reported on the kernel bug tracker | 18:03 |
mth | yes, sounds like a good idea | 18:03 |
Ayla | at which point is it better to use msleep over mdelay, or the other way around? | 18:08 |
Ayla | on some parts of the code, I waits only 10ms | 18:09 |
Ayla | s/I/it | 18:09 |
Ayla | msleep() probably introduces a higher delay | 18:09 |
wpwrak | "only" 10 ms :-) | 18:10 |
Ayla | but I'm not sure it's important whether it waits longer than 10ms or not | 18:10 |
wpwrak | that's a pretty nasty delay, if you spend that time spinning | 18:10 |
Ayla | well, the panel enable function spins for like one second total | 18:11 |
Ayla | 510ms for the ili9325 panel | 18:12 |
wpwrak | mdelay ? urgh | 18:12 |
Ayla | yes :) | 18:12 |
Ayla | IIRC we use HZ=250 on dingoo, so a msleep() will last 4ms minimum | 18:13 |
Ayla | and msleep(10) will sleep 12 milliseconds | 18:13 |
wpwrak | just hope you don't have anything remotely real-time-ish on that machine :) | 18:14 |
Ayla | we do, sort of | 18:15 |
Ayla | the sound output | 18:15 |
larsc | which is taken care of by dma | 18:17 |
mth | only for the span of the current period though | 18:18 |
mth | the sound output is broken when the SLCD init happens | 18:18 |
mth | when unblanking, for example | 18:19 |
mth | Ayla: I think that even 1 ms is a lot of time for this CPU | 18:19 |
Ayla | larsc: yes, but the app that produces sound is blocked | 18:19 |
mth | and afaik the 4 ms time slices matter when there are multiple active threads, but it doesn't mean a single active thread has to wait for the next time slice | 18:20 |
Ayla | ok | 18:20 |
Ayla | I replaced all mdelays by msleeps, and it works good | 18:20 |
Ayla | now the sound doesn't stop when I un-blank the screen on GMU | 18:20 |
mth | zLoud will be so happy | 18:21 |
Ayla | heheh | 18:22 |
qi-bot | [commit] Paul Cercueil: MIPS: JZ4740: Delay using msleep to prevent spinning for 0.5s. (jz-3.4) http://qi-hw.com/p/qi-kernel/b5cb662 | 18:25 |
Ayla | mth: could you mark it as done on the bug tracker? | 18:28 |
mth | yep, works great | 18:38 |
mth | issue closed | 18:39 |
qi-bot | [commit] Maarten ter Huurne: MIPS: JZ4740: Work around fbcon logo crash with SLCD. (jz-3.4) http://qi-hw.com/p/qi-kernel/d072b59 | 19:22 |
whitequark | viric: where did you get your looooongson? | 21:34 |
lekernel | by the way, what's the status of the kicad patches? were they submitted upstream? | 21:35 |
wpwrak | lekernel: at some point they were, and received the reply that this wasn't a good time | 22:46 |
GNUtoo-desktop | viric, how's the lemote? | 22:52 |
--- Sun Jun 17 2012 | 00:00 |
Generated by irclog2html.py 2.9.2 by Marius Gedminas - find it at mg.pov.lt!