**** BEGIN LOGGING AT Mon Aug 15 02:59:57 2011 Aug 15 08:41:32 hi bluelightning Aug 15 08:41:41 ant_home: morning Aug 15 08:44:17 ant_home: what did you mean by disastrous? Aug 15 08:49:34 no prompt :/ Aug 15 08:50:02 I'm booting now core-image-minimal Aug 15 08:52:01 err.. trying to boot :p Aug 15 08:52:09 no life Aug 15 09:01:10 hmm... I managed to boot 3.0 on my hx4700 without too much trouble over the weekend Aug 15 09:03:50 (and no trouble with the touchscreen on that machine BTW) Aug 15 09:53:49 sorry, I'm playing with the kid in the meantime... Aug 15 09:54:09 ok, I could build core-image-sato Aug 15 09:54:19 with two adjustments Aug 15 09:54:32 one: oe-core lacks provider for lzma-native Aug 15 09:54:52 two: image_types.bbclass sould set ?= the JFFS2 stuff Aug 15 09:55:11 so the settings from zaurus.inc have chanche to go Aug 15 09:55:36 ERROR: Nothing PROVIDES 'lzma-native' Aug 15 09:55:46 ^^ this is provided by xz nowadays Aug 15 10:02:27 ah, missing patch Aug 15 10:02:32 is already in oe-dev Aug 15 10:10:49 fix sent to JaMa Aug 15 21:31:54 lumag: welcome here :) Aug 15 21:32:00 ok, this is better :) Aug 15 21:32:26 lumag: btw, we have strange bug in kexecboot.. Aug 15 21:32:46 for first time one pointer is NULL on arm but I've not seen that bug on x86 Aug 15 21:33:02 which pointer? Aug 15 21:33:07 I'll show the code Aug 15 21:34:04 http://git.linuxtogo.org/?p=groups/kexecboot/kexecboot.git;a=blob;f=kexecboot.c;h=b0ad99979bc75337fbcd7da64ee979b9e317ce2e;hb=447411662de0a4ba148a816c155018488b082c35#l722 Aug 15 21:34:45 here menu->current is NULL for first time you entering this menu level Aug 15 21:34:57 well.. it's hard to understand w/o background Aug 15 21:35:07 but I'm sure it shouldn't be NULL :) Aug 15 21:35:16 because it is always initialized Aug 15 21:35:24 how better to debug this case? Aug 15 21:35:45 it's almost always null on pxa250, rarely on pxa255, pxa270 Aug 15 21:35:48 http://git.linuxtogo.org/?p=groups/kexecboot/kexecboot.git;a=commitdiff;h=3da0d9203c0b705feaa0bc0ef01ebd22e742d48a;hp=447411662de0a4ba148a816c155018488b082c35 Aug 15 21:36:02 here I've added some code to debug Aug 15 21:36:07 Jay7: remember it looks like a 'double click' issue Aug 15 21:36:37 if you carefully press System Menu you have more chanches to enter Aug 15 21:36:40 ant__: which contition we fall into? Aug 15 21:36:44 *condition Aug 15 21:36:50 if you kjeep pressed is almost surely freeze Aug 15 21:36:54 menu->current or menu->current->current? Aug 15 21:37:01 2nd Aug 15 21:37:18 I see only one way to debug this Aug 15 21:37:20 Did you try valgrinding it? Aug 15 21:37:26 use rootfs with gdb :) Aug 15 21:37:37 I have no ARM devel env.. Aug 15 21:37:45 lumag: I noticed an alignment trap Aug 15 21:37:53 we debugged and come to this code Aug 15 21:37:54 OE have sdk image but it is broken constantly Aug 15 21:38:04 wrong perl things in autotools Aug 15 21:38:39 but I see no way to debug it with eyes :) I've tried and see nothing wrong Aug 15 21:38:44 on x86 all is fine Aug 15 21:38:50 valgrind on x86? Aug 15 21:38:59 yes Aug 15 21:39:03 no problems Aug 15 21:39:10 strange Aug 15 21:39:52 well, np, I'll spend some time on this later Aug 15 21:39:59 Anyway let me first build a usable kexecboot for tosa, so that I can test anything :) Aug 15 21:40:10 hehe.. enter into System menu :) Aug 15 21:40:17 it should hang Aug 15 21:40:31 but it may not hang.. Aug 15 21:40:40 kind of floating bug Aug 15 21:40:53 but we are not using threads :) Aug 15 21:41:25 fwiw running kexecboot as binary from SD allowed to take the logs of 'alignment trap' and did not freeze iirc Aug 15 21:41:53 strace'ing it may be good idea.. Aug 15 21:42:00 it looks like the alignment is fatal when kexecboot runs in the initramfs Aug 15 21:42:08 JAy, it's null.. Aug 15 21:42:09 but iirc, strace will show syscalls mainly Aug 15 21:42:36 anyway, I need gdb on ARM rootfs with autotools Aug 15 21:42:57 http://paste.debian.net/126341/ Aug 15 21:43:02 ^^ patch + logs Aug 15 21:43:35 first time you enter is null Aug 15 21:43:52 If you can run kexecboot from SD, do you have such problem with glibc image? Aug 15 21:44:42 you mean kexecboot-eglibc? iirc we did this also Aug 15 21:45:01 but the klibc was different..libs are statically linked Aug 15 21:45:15 so debug was straightforward (almost :) Aug 15 21:45:36 now, with the a.m. hack we filter the NULL case Aug 15 21:45:45 and I'd say, push it for the moment Aug 15 21:46:12 better than freeze Aug 15 21:47:04 Jay7: and the 0001-Check-devtmpfs-presence-in-runtime.patch too Aug 15 21:47:10 :) Aug 15 21:48:42 lumag: another subtle thing..about event polling rate. Did it change with 2.6.3x Aug 15 21:48:46 ? Aug 15 21:49:38 e.g. locomo patch does Aug 15 21:49:40 -#define SCAN_INTERVAL<><------>(HZ/10) Aug 15 21:49:40 +#define SCAN_INTERVAL<><------>(HZ/20) Aug 15 21:49:40 . Aug 15 21:50:30 ant__, I saw that there was major change in keyboard handling. I don't know about polling rate. Aug 15 21:51:12 hm.. may be we hit this Aug 15 21:51:30 or I've just broken event handling :) Aug 15 21:52:22 seems all slowish Aug 15 21:52:32 2.6.2x was 'snappy' Aug 15 21:53:00 but Yuri's patch surely has some side-effect ;p Aug 15 21:53:30 yeah, as always ;) Aug 15 21:53:40 I told him not to optimize too much... ;) Aug 15 21:53:49 we risk gcc issues :p Aug 15 21:53:58 :) Aug 15 21:54:40 I didn't check the code. But basically IIRC userspace has no control on the speed of the keyboard, if it does just poll/select on event devices. Aug 15 21:55:53 there is ioctl to change rate Aug 15 21:56:10 we are using it if enabled at configure stage Aug 15 21:57:43 I see Aug 15 22:16:35 this is the oe-core-sato log Aug 15 22:16:38 http://paste.debian.net/126345/ Aug 15 22:29:11 * Jay7 -> sleep Aug 15 22:36:40 gn Jay7 Aug 15 22:54:51 lumag: strangely enough I can't reproduce the a.m. NULL case ... Aug 15 22:55:13 I just applied the keymap patch Aug 15 22:55:22 ah, and built with gcc4.6 Aug 15 23:34:39 gn **** ENDING LOGGING AT Tue Aug 16 02:59:57 2011