**** BEGIN LOGGING AT Wed Feb 10 02:59:58 2010 Feb 10 03:11:54 zhenhua: yes, I got the segfault here. Feb 10 03:12:29 padovan: ok. i think you have got what i means. we need to check and return if data is null Feb 10 03:19:57 zhenhua: we can only check for data. It's null on this case. Feb 10 03:25:58 yes Feb 10 03:31:10 zhenhua: we need to move _dbus_register_interface down or unregister the interface on hfp_probe. Feb 10 03:31:28 Actually, that is another bug :) Feb 10 03:56:57 how is your modemconf even creating the hfp modem Feb 10 03:58:25 I suspect that if no helper is defined it should never register the modem Feb 10 03:58:25 parse_config create a modem for each keyfile Feb 10 04:07:16 padovan: any idea to fix the issue? i don't quite catch it? Feb 10 04:07:34 i just back after my lunch. :-) Feb 10 04:11:40 zhenhua: in hfp_probe move g_dbus_register_interface to after if (hfp_register_ofono_handsfree(modem) != 0) Feb 10 04:12:04 and in hfp_register_ofono_handsfree add if (!data) return -EINVAL; Feb 10 04:12:56 why not simply return EINVAL in probe? Feb 10 04:14:11 denkenz: good idea, just need to add a ofono_modem_get_data there. Feb 10 04:15:48 sounds good to me Feb 10 04:19:52 padovan: why we need a function hfp_load_modems(). i would simply put it into hfp_init Feb 10 04:21:50 zhenhua: we can change it. That's ok for me. Feb 10 04:22:40 padovan: thanks. Feb 10 04:28:16 padovan: in list_devices_cb, i doubt we should use g_strfreev(device_list) instead of g_free Feb 10 04:30:08 padovan: same as list_adapters_cb Feb 10 04:31:09 it's a bug.:-) Feb 10 04:32:25 nope, look closer Feb 10 04:33:08 no? Feb 10 04:33:38 no, it is actually correct Feb 10 04:33:58 i means in the done: Feb 10 04:34:20 done: + g_free(device); dbus_message_unref(reply); Feb 10 04:34:42 sorry, g_free(device_list) Feb 10 04:34:49 it's char ** Feb 10 04:35:20 it is, but again, look closer into where the device_list[i] goes Feb 10 04:35:47 hmm... i got it now. Feb 10 04:36:06 zhenhua: the g_strfreev(adapter_list); after done: label is right. Feb 10 04:36:21 we are leaking there. Feb 10 04:37:06 i am talking about list_devices_cb Feb 10 04:37:31 should g_free(device_list) should above done label? Feb 10 04:38:11 anyway, it's ok to free a null and g_free simply returns. Feb 10 04:55:31 hey, getting a build error with latest git: Feb 10 04:55:34 gisi/netlink.c: In function ‘g_pn_nl_process’: Feb 10 04:55:34 gisi/netlink.c:97: error: cast increases required alignment of target type Feb 10 04:55:34 gisi/netlink.c:129: error: cast increases required alignment of target type Feb 10 04:55:34 gisi/netlink.c:130: error: cast increases required alignment of target type Feb 10 05:01:58 what's your compile option? i don't meet that Feb 10 05:12:47 ./bootstrap-configure ; make Feb 10 05:13:22 http://www.mail-archive.com/ofono@ofono.org/msg00621.html Feb 10 05:18:21 jebba: this keeps popping up from time to time Feb 10 05:18:35 I honestly don't remember how to solve it other than getting a new compiler or turning off isi Feb 10 05:21:29 well, i need isi. I just do ./configure without "--enable-maintainer-mode" to work around it for now. denkenz Feb 10 05:21:35 what's it casting? Feb 10 05:21:38 and why Feb 10 05:21:49 maybe there's a gnu flag or something? Feb 10 05:22:07 tmzt: check that thread, it 'splains it Feb 10 05:22:16 something linux linux's containerof? Feb 10 05:22:20 hmm Feb 10 05:23:03 jebba: it doesn't show up on make distcheck, so I suspect compiler versions Feb 10 05:23:57 gcc (GCC) 4.4.2 20091019 (Red Hat 4.4.2-5) Fedora 12 Feb 10 05:25:43 then we need to pragma this stuff out Feb 10 06:18:20 denkenz: still there? Feb 10 06:19:21 denkenz: in g_at_chat_cleanup, why we set chat->channel to NULL instead of g_io_channel_shutdown() then unref(). Feb 10 06:23:08 denkenz: and then set chat->channel to NULL. Feb 10 06:27:24 oh, apologize for my stupid. when the destroy notify func is called, the io channel is already shutdown. is that right? Feb 10 06:30:59 we don't unref the channel, it is done when read watch is destroyed Feb 10 06:31:22 this is why we set close on unref flag in the beginning Feb 10 06:33:14 got u Feb 10 06:42:18 $ sudo ./test/enable-modem Feb 10 06:42:19 Traceback (most recent call last): Feb 10 06:42:19 File "./test/enable-modem", line 14, in Feb 10 06:42:19 path = properties["Modems"][0] Feb 10 06:42:19 IndexError: list index out of range Feb 10 06:42:59 I'm running `ofonod -n -d 99` as root. This on N900 under fedora 12. It's not giving any debug info, which i find different from last time i did it. Feb 10 06:45:43 looks like modem is not detected by ofono Feb 10 06:46:14 you can run ofonod -n -d '*' to output all debug info Feb 10 06:46:57 what's u kernel version? mine is 2.6.31-16 on ubuntu Feb 10 06:47:41 2.6.28-omap1-jebba10-g5a58882 (basically same as Maemo N900 kernel, with a few more things enabled like framebuffer console) Feb 10 06:47:44 it needs phonet driver to detect N900 I believe. Feb 10 06:48:19 ya, phonet driver is loaded. I have gotten this working with N900 under maemo (& phonet). Now i'm trying to do the same under fedora. Feb 10 06:48:26 you're running ofono on your desktop instead of maemo phone, right? Feb 10 06:48:46 connect to N900 by usb? Feb 10 06:49:06 then u need to ask N900 to export himself to kernel Feb 10 06:49:16 s/himself/itself Feb 10 06:49:41 no, in this case I am running everything directly on the phone. Feb 10 06:50:24 u rebuild N900 with fedora? Feb 10 06:52:02 i can't help here because isimodem is different story with atmodem. Feb 10 06:52:04 N900 is the fone itself, the OS that it comes with is "maemo". What I have is fedora 12 built for arm running maemo/n900 kernel. Feb 10 06:52:50 i was able to get isimodem working with n900 under maemo (without connecting PC). Feb 10 06:54:23 check ofono/gisi/netlink.c g_pn_netlink_start() Feb 10 06:54:52 isimodem_init() call it to get the link Feb 10 06:55:49 there's a fd coming out from kernel phonet driver. I guess. Feb 10 06:55:52 fd = socket(PF_NETLINK, SOCK_DGRAM, NETLINK_ROUTE); Feb 10 09:55:15 ofonod[761]: drivers/atmodem/sim.c:at_crsm_info_cb() crsm_info_cb: 90, 00, 36 Feb 10 09:55:15 ofonod[761]: Requested file structure differs from SIM: 6fb7 Feb 10 09:56:04 ofonod[761]: drivers/atmodem/sim.c:at_crsm_info_cb() crsm_info_cb: 90, 00, 36 Feb 10 09:56:04 ofonod[761]: Requested file structure differs from SIM: 6fb7 Feb 10 14:42:37 yeah that part is the weird way we request EFecc Feb 10 14:42:57 ignore that error for now, we try both file structures Feb 10 15:56:17 we could try only the one hinted by the sim phase, Feb 10 15:56:43 but there wouldn't be any gain performance wise Feb 10 16:00:27 would squash that warning ;) Feb 10 19:01:15 denkenz: ping Feb 10 19:02:09 jprvita: pong Feb 10 19:02:34 denkenz: what do you use signalfd on ofono for? Feb 10 19:03:08 safe signal handling, in particular exiting the event loop Feb 10 19:03:08 I'm trying to cross-compile it on a arm board, but I have libc 2.5 only there :( Feb 10 19:03:47 do you think it's easy to patch ofono to drop this dependency? Feb 10 19:04:12 Well we specifically switched to signalfd Feb 10 19:04:58 I mean, I don't want to upstream it, just maintain a patch for this particular purpose Feb 10 19:05:16 ah, yes it is possible Feb 10 19:05:25 at least until I don't have a newer libc on the sdk i'm working with Feb 10 19:05:28 just git log main.c and see how it was before Feb 10 19:06:07 cool, tks! Feb 10 19:09:43 it seems I just need to revert adfb6c9d: Use signalfd for signal handling Feb 10 19:09:48 let's try Feb 10 22:16:23 glibc probably even has an implementation of it for old kernels that you can copy Feb 11 01:12:53 on fedora 12 arm with nokia n900, running this first was able to get ofonod/isimodem working: Feb 11 01:12:53 /mnt/initfs/sbin/phonet -a 0x6C -i phonet0 **** ENDING LOGGING AT Thu Feb 11 02:59:57 2010