**** BEGIN LOGGING AT Fri Oct 01 02:59:58 2010 Oct 01 05:01:08 pessi, akiniemi: What is usbpnmodem.c still there for? Oct 01 05:01:14 We don't even compile that one right now. Oct 01 05:05:18 Can we please remove that one. Oct 01 05:05:48 Or enable it to build it again. Oct 01 06:43:10 holtmann: hi, yes, I think it was supposed to be removed with the "Remove usbpbmodem plugin" patch Oct 01 06:43:19 holtmann: I'll ping aki as well Oct 01 06:43:48 kvehmanen: So we are having AudioSettings interface now that can be used to interact with cmt_speech. Oct 01 06:43:57 What do you actually need from us for real. Oct 01 06:44:03 holtmann: I was just about to ask about that :) Oct 01 06:44:07 Or is that just about PA. Oct 01 06:44:45 holtmann: I was actually supposed to send a patch that adds something similar but you made it before me ;) Oct 01 06:45:06 We most likely still need a patch for the ISI modem. Oct 01 06:45:08 holtmann: but in short, looks good to me and I'll make a patch for isimodems to use this Oct 01 06:45:11 I only did Huawei and IFX. Oct 01 06:45:41 I also wanna include Device=/dev/cmt_speech or whatever you guys use. Oct 01 06:46:03 For Huawei this is important when you stick two or more dongles into one machine. Oct 01 06:46:36 For N900 maybe not so much. Oct 01 06:46:47 holtmann: so in current meego appraoch, the client for this API is at http://meego.gitorious.org/maemo-multimedia/pulseaudio-modules-meego Oct 01 06:47:51 holtmann: this is a module loaded to into pulseaudio that does all the magic for voice calls (including voice processing, routing and enforcement point for policy actions) Oct 01 06:48:07 Can we use that for Huawei as well? Oct 01 06:48:21 There we just have 8khz 16-bit 20ms period PCM stream. Oct 01 06:48:27 holtmann: for N900, this module uses libcmtspeechdata to get the speech frames which again uses /dev/cmt_speech) Oct 01 06:49:05 holtmann: for sure yes, but some work will be needed as the modem interface is different Oct 01 06:49:13 I guess Huawei has a ALSA interface? Oct 01 06:49:33 holtmann: but AFAIK the ifx interface is different as well, so this is just business as usual Oct 01 06:49:51 kvehmanen: No. Only a TTY :( Oct 01 06:50:29 holtmann: the idea current is that pulseaudio-modules-meego has a common voice module that is common for all modems (plus VoIP calls), and then there will be a separate PA module for each different modem speech API Oct 01 06:51:07 holtmann: so it sounds like huawei would needs its own PA module Oct 01 06:51:07 Okay. So my knowledge about PA is limited. Oct 01 06:51:38 If anybody wants to take a stab at it that would be great. Since with the two Huawei voice capable dongles you can then test things on a normal PC. Oct 01 06:51:42 holtmann: note that this is not yet in upstream PA, but pulseaudio-modules-meego is still a separate project Oct 01 06:52:18 However if every modem is that different, we might just need to add a Type or Vendor property as well :( Oct 01 06:52:26 holtmann: but e.g. work is ongoing to adapt the package to handle both ifx and nokiamodem, so it would make sense to add the huwwei case as well Oct 01 06:52:36 That would be nice. Oct 01 06:53:00 And can we please call the module with an ofono prefix for the cases it talks to ofono. oFono != MeeGo. Oct 01 06:53:01 holtmann: some generic info abou the arch -> http://linuxplumbersconf.org/2009/slides/Jyri-Sarha-audio_miniconf_slides.pdf Oct 01 06:53:21 I am still not an audio guru :( Oct 01 06:54:07 holtmann: I wouldn't worry about that.. I think the goal is to get this (or major parts) of this to upstream PA and then it would have no inherent meego connection Oct 01 06:54:41 It if takes to oFono it should be called ofono. That is just my point here. Oct 01 06:55:49 holtmann: that's of course interesting discussion, whether some of the adaptation should be done in ofono context? Oct 01 06:56:21 So you need to detect PCM on/off from oFono. Plus you need the basic settings from oFono. Like burstmode etc. Oct 01 06:56:28 At least that is what I've been told. Oct 01 06:56:31 holtmann: e.g. stuff like libcmtspeechdata that absracts the modem speech interface from Pulseaudio Oct 01 06:57:19 So currently we have udev configuration for IFX that set the things like burstmode depending on its udev config and then exports that back over the AudioSettings API. Oct 01 06:59:04 holtmann: for nokia case, we just need a boolean on/off toggle indicating that voice call service is active Oct 01 06:59:07 So in theory the PA module that talks to oFono, should get these and then configure cmtspeech or just deal with Huawei audio TTY or something else. Oct 01 06:59:29 For the IFX it seems the burstmode/fullduplex setting is also needed. Oct 01 06:59:37 holtmann: libcmtspeechdata uses this info to set up (and close) the physical speech interface towards the modem Oct 01 07:00:28 And Huawei just gives the khz, bitrate and ms period of its PCM stream. It does inband ring generation anyway. So there the modem does everything for you. Oct 01 07:00:35 holtmann: for the nokiamodem case, all the rest of the configuration is negoatiated over the /dev/cmt_speech link (there's a ISI-style miniprotocol to talk to the modem DSP) Oct 01 07:01:05 Do you happen to know if that is also the case for IFX and we are just complicating things? Oct 01 07:01:15 Or there we need more information. Oct 01 07:01:27 huawei and ifx cases probably both need more help to set up the channel Oct 01 07:01:37 Fair enough then. Oct 01 07:01:53 So if we write a simple ifxmodem/audio-settings.c support you guys should be good. Oct 01 07:02:00 s/we/you/ Oct 01 07:03:00 holtmann: but I think the current interface is a great starting point, and any extensions should be done as the need become clear (from folks knowning the modem interface) Oct 01 07:03:45 I prefer that as well. I am running blind here btw. Oct 01 07:04:21 holtmann: I've understood that also ifx has methods to signal some stuff inband, so in the end the current simple interface might just be enough Oct 01 07:04:55 holtmann: on nokia side this interface has a funny history, as after a lot of discussion and proposals we ended up with essentially just a one-bit protocol Oct 01 07:05:14 and thus in n900, the interface is now better known as "pekka's one-bit protocol" ;) Oct 01 07:05:44 Hah. Oct 01 07:07:39 holtmann: but yes, I'll first complete the n900 support for the interface (and try to send my first ofono patch), and will then try to ping someone knowing the ifx stuff on how the current interface looks for them Oct 01 07:08:47 Sounds good to me. Oct 01 07:09:08 If you make the isimodem patch within the next 1-2 hours, you might get it into ofono-0.31 Oct 01 07:09:14 I will release that one today. Oct 01 07:11:50 holtmann: aa, I'll try but no promises.. I'm currently wondering how to test this with some version of PA on n900 :P Oct 01 07:12:26 You can just run something like monitor-ofono python script to see the D-Bus signals. Oct 01 07:12:39 Or just list-modem script to see current values. Oct 01 07:15:05 holtmann: yep, done that actually, but would like to see it working with an actual call on a real device (hearing is believing) Oct 01 07:15:28 Fair enough. Oct 01 07:23:41 holtmann: usbpnmodem.c needs to be removed. Oops. Oct 01 07:28:19 akiniemi: I do that then. No problem. Oct 01 07:29:07 Done. Oct 01 07:29:42 akiniemi: I had to do some cleanups. Oct 01 07:30:07 Can you guys please use the copyright and license header that every other file uses and just replace the copyright line. Oct 01 07:30:15 I want consistency here. Oct 01 07:31:12 holtmann: Sure, I can do that. Oct 01 07:31:45 I fixed plugins directory now. Oct 01 07:31:54 Not that we always get this right either btw. Oct 01 07:32:08 Found us slagging off also a few times. Oct 01 07:34:54 holtmann: crap. I see some of Pekka's patches that I pushed still have the wrong committer email address. Oct 01 07:35:01 I thought I had fixed that :( Oct 01 07:35:15 holtmann: Do you want me to revert&fix? Oct 01 07:35:24 Nope. Too late. Oct 01 07:35:53 Sigh. Well at least the display name is right... Oct 01 07:36:28 I fix it up via .mailmap for you. Oct 01 07:37:49 His display name was also messed up in earlier commits. Oct 01 07:38:58 akiniemi: Push now. So at least git shortlog does the right thing. Oct 01 07:39:49 s/Push/Pushed/ Oct 01 07:55:26 holtmann: thanks Oct 01 08:04:21 No problem. Aren't the first ones :( Oct 01 09:58:22 I have a Huawei ec1260 modem. It gets detected without installing usb_modeswitch on my MeeGo netbook. How do I add support for it to ofono? Oct 01 09:58:31 I followed http://ofono.org/wiki/how-enable-modem-ofono . Oct 01 09:59:10 I dont know what to add in the drivers field. Oct 01 10:00:32 Actually I started ofono in debug mode, and when I plugin the modem it prints 3 lines. src/modem.c : ofono_modem_create() name: ???????????? type:huawei Oct 01 10:24:03 debayan: what ofono version? Oct 01 12:30:09 btw, what's the policy with git commit signed-offs... are they optional, or actually recommended against? Oct 01 12:30:29 We are not using them. Oct 01 12:30:59 holtmann: and oops, coding-style.txt actually does mention this... sorry for the noise Oct 01 12:40:33 No problem. Oct 01 13:24:19 holtmann: is there a reason not to use udev_device_get_property_value () and go through the property entries explicitly? Oct 01 13:24:41 pessi: Most likely, because that was not there yet. Oct 01 13:25:09 If you look for more than one value, it might be not efficient. Not that this would really matter. Oct 01 13:27:06 nope, just wondering Oct 01 13:27:52 Me being stupid is also a possibility. Oct 01 13:28:35 env{}+= stuff is not b Oct 01 13:28:43 ...is not there multiple times Oct 01 13:40:24 143 does not have udev_device_get_property_value, that is why ;) Oct 01 13:40:34 146 does not either Oct 01 14:04:57 So I am not fully braindead yet ;) **** ENDING LOGGING AT Sat Oct 02 02:59:57 2010