**** BEGIN LOGGING AT Thu Jul 01 02:59:57 2010
Jul 01 14:59:17 <denkenz> holtmann: Are you fine with Zhenhua's em770 removal patches?
Jul 01 15:01:14 <holtmann> Do we really wanna do that?
Jul 01 15:01:23 <holtmann> What is the motivation behind it.
Jul 01 15:01:36 <holtmann> I did like the idea to keep dongles apart from voice capable modems.
Jul 01 15:02:29 <denkenz> The only difference between em770 is the addition of voicecalls
Jul 01 15:02:40 <denkenz> Although EM770 is also capable of high-speed with some kernel driver hacks
Jul 01 15:02:48 <denkenz> Nobody managed to do that though yet
Jul 01 15:03:18 <holtmann> Let me have a second look at that patch.
Jul 01 15:03:29 <denkenz> So it seems silly to maintain 2 files that are almost completely the same
Jul 01 15:04:22 <holtmann> Fair enought.
Jul 01 15:04:27 <holtmann> So I looked through it again.
Jul 01 15:04:40 <holtmann> One problem I have with this patch is that we check the USB PID inside the source code.
Jul 01 15:04:52 <holtmann> We should have a udev env for giving us the model.
Jul 01 15:05:08 <holtmann> Something like OFONO_HUAWEI_VOICE=yes/no.
Jul 01 15:05:16 <denkenz> Yeah I don't like that part too
Jul 01 15:05:39 <holtmann> Then let me reply to it. And have him fix that.
Jul 01 15:05:50 <holtmann> Otherwise fine with me. The less files the better.
Jul 01 15:07:09 <denkenz> Sounds good
Jul 01 15:09:27 <holtmann> Sent.
Jul 01 15:26:18 <asac_> holtmann: hey sir. do you have suggestion on hardware we should use/choose to get a voice call experiment/experience going on top of ofono?
Jul 01 15:27:00 <asac_> at best something more or less easily available to community :)
Jul 01 15:27:06 <denkenz> Telit hardware or Freerunner
Jul 01 15:27:16 <denkenz> I have UC864 and its fairly nice
Jul 01 15:27:28 <asac_> denkenz: hey. let me google it ;)
Jul 01 15:28:09 <asac_> denkenz: is that supported in linux oob? how does sound work? did you need to wire something up?
Jul 01 15:28:31 <denkenz> Sound works using a 2.5mm headphone or you can wire an external speaker
Jul 01 15:28:53 <denkenz> And it works pretty much out of the box
Jul 01 15:29:05 <denkenz> You might need to tweak the serial id for usb_serial to recognize it
Jul 01 15:29:12 <asac_> yeah
Jul 01 15:29:25 <asac_> denkenz: 2.5mm headphone? isnt that an internal module?
Jul 01 15:29:48 <denkenz> It comes with a dev board
Jul 01 15:30:11 <denkenz> You can run it without one, but that's more than my soldering skills can handle
Jul 01 15:30:42 <asac_> oh
Jul 01 15:30:48 <asac_> denkenz: is that an intel or arm devboard?
Jul 01 15:30:55 <asac_> where can i buy that ;)?
Jul 01 15:31:07 <denkenz> I have no idea, don't think it has a CPU
Jul 01 15:31:13 <denkenz> just USB + antenna
Jul 01 15:31:36 <denkenz> And you can buy it at the regular places, like digi key, etc
Jul 01 15:31:48 <denkenz> Telit has a list of dealers on their website
Jul 01 15:32:22 <denkenz> The only other alternatives are Freerunner or hacked phones like the G1
Jul 01 15:32:37 <denkenz> I don't know if anyone got the G1 fully working though
Jul 01 15:32:48 <denkenz> Oh, and the N900 of course ;)
Jul 01 15:33:51 <asac_> great thanks for the info!!
Jul 01 15:33:57 <asac_> i will checkout telit or get N900 ;)
Jul 01 15:34:28 <asac_> price of the telit evaluation kit seems to be close to N900 ;)
Jul 01 15:35:08 <denkenz> yeah its a bit pricey
Jul 01 15:35:25 <denkenz> but you get full docs
Jul 01 15:36:39 <denkenz> The only other lead I can give you is a TI Omap Zoom or Zoom II
Jul 01 15:36:52 <denkenz> But those are even more $ and rather uncommon
Jul 01 15:51:38 <pessi> denkenz: the dtmf tones?
Jul 01 15:51:54 <denkenz> ?
Jul 01 15:52:18 <pessi> is it possible to have a separate atom for them, if and when modem supports some kind of interfce to senddtmf and stopdtmf?
Jul 01 15:52:53 <denkenz> yes, its called custom driver ;)
Jul 01 15:54:05 <denkenz> But then you have trouble if you want to support pauses and dtmf strings inside voice calls
Jul 01 16:10:16 <lizardo> denkenz: is ofono already working on N900 ? last time I tried (several months ago) it was not there yet... Did things changed lately?
Jul 01 16:10:57 <denkenz> I'm the wrong person to ask, last I checked if oFono runs with csd then things just work magically
Jul 01 16:11:09 <denkenz> Not sure if anyone got oFono working standalone with voice on it
Jul 01 16:12:39 <lizardo> denkenz: yeah, I meant standalone :) will try again some day...
Jul 01 16:38:22 <pessi> denkenz: ...or signal telling what we have left from SendTones?
Jul 01 16:41:31 <pessi> lizardo: I had the native N900 patch, I suppose it needs some disclaimers so you know what you do when you fry your sim card ;-)
Jul 01 16:45:15 <denkenz> pessi: I'm open to suggestions here, the DTMF is a really nasty part of the spec
Jul 01 16:45:39 <denkenz> But we need to make it work on 27.007 style hw 
Jul 01 16:50:05 <pessi> denkenz: signal SendingDTMF to VoiceCallManager?
Jul 01 16:50:36 <denkenz> What does that accomplish? And how would you do it on an AT modem?
Jul 01 16:55:07 <pessi> 1) give some feedback when DTMFs are being sent and 2) wait until each respective at+vts returns OK?
Jul 01 16:57:34 <denkenz> Actually the spec makes no guarantees that getting a VTS OK is the same as stop tone
Jul 01 16:59:15 <denkenz> And if you decide to use pauses, then you're even worse off
Jul 01 16:59:35 <denkenz> Since you either need to manage them manually or rely on ATD
Jul 01 16:59:43 <denkenz> Which gives _no_ feedback on DTMF tones
Jul 01 17:00:13 <denkenz> The whole DTMF + pause stuff is a legacy insanity
Jul 01 17:06:13 <pessi> well, that is what we are doing, legacy insanity ;)
Jul 01 17:06:55 <lizardo> pessi: would the ISI "firewall" allow you to fry the SIM card ? 
Jul 01 17:07:05 <denkenz> yeah, but there are different levels of insanity
Jul 01 17:07:14 <denkenz> DTMF is pretty much the highest level
Jul 01 17:07:26 <denkenz> STK call control is slightly below it
Jul 01 17:07:29 <pessi> I'd rather got to some solution which may work with some modems, and use quirks with offending ones
Jul 01 17:08:09 <denkenz> We do have a solution, it just doesn't satisfy you ;)
Jul 01 17:08:35 <pessi> lizardo: the gpio lines allow you to reset modem at inconvenient moments
Jul 01 17:09:45 <pessi> denkenz: it is just a solution which does not my solve problems
Jul 01 17:10:48 <denkenz> I know, but my point here is that there is no generic solution
Jul 01 17:12:03 <denkenz> If there is no generic solution, then you have to solve it at the platform / driver level until we find one
Jul 01 17:13:11 <pessi> I'd be happy with high-level interface so that we could start doing it at driver level
Jul 01 17:13:44 <denkenz> How can we design a high level interface if the hardware does not support it?
Jul 01 17:13:57 <denkenz> Faking it ala telepathy is not going to happen
Jul 01 17:17:18 <denkenz> The first place I'd start is checking whether modem -> network signaling even supports durations
Jul 01 17:17:27 <denkenz> I suspect they're all fixed duration anyway
Jul 01 17:17:31 <pessi> nope
Jul 01 17:17:47 <pessi> there is ISDN message SendDTMF which you give the tone you want to send
Jul 01 17:18:07 <pessi> after appropriate delay, you send StopDTMF
Jul 01 17:18:15 <pessi> sorry, StartDTMF/StopDTMF
Jul 01 17:19:10 <pessi> isimodem and maemo csd expose them
Jul 01 17:19:40 <denkenz> Ok fair enough
Jul 01 17:19:47 <pessi> but of course, if you use at commands, you are mostly happy with atd ot at+vts
Jul 01 17:19:56 <denkenz> However, AT ones don't, the best you can do is use the extended VTS form
Jul 01 17:20:04 <denkenz> where you specify the duration
Jul 01 17:20:13 <denkenz> So doing start / stop tone is simply not possible
Jul 01 17:24:11 <pessi> yep.. so we have to fake it in ring
Jul 01 17:24:56 <denkenz> Or you need to redesign that API
Jul 01 17:25:07 <denkenz> But that's not my bug :)
Jul 01 17:25:49 <pessi> i added a proprietary dialstring upport to tp-ring
Jul 01 17:28:19 <pessi> and there seems to be ...MultipleTones added to 0.19.6
Jul 01 17:28:55 <pessi> whee
Jul 01 17:33:04 <denkenz> good luck with that, RTP and GSM don't mix ;)
Jul 01 18:58:00 <kvalo> denkenz: btw, huawei PIN unlock is broken. I think my SYSINFO patches broke it :(
Jul 01 19:05:40 <holtmann> kvalo: Then you have to fix it ;)
Jul 01 19:18:03 <denkenz> kvalo: Define broken? Log?
Jul 01 19:19:33 <kvalo> holtmann: heh. my vacation starts tomorrow, ofono has to wait :)
Jul 01 19:21:09 <kvalo> denkenz: I don't have logs here. but it was so something like SYSINFO test never is true if SIM is locked
Jul 01 19:21:49 <kvalo> I found some documentation which says that the variable we check actually means that SIM is not ready or SIM is locked
Jul 01 19:22:37 <denkenz> ah ok, so your sim inserted test never works
Jul 01 19:22:46 <denkenz> we ran into the same with mbm ;)
Jul 01 19:23:37 <kvalo> why can't they use separate value for sim not ready or locked?
Jul 01 19:23:52 <denkenz> Because that would make sense
Jul 01 19:23:58 <denkenz> Can't have that
Jul 01 19:24:01 <kvalo> heh
Jul 01 22:43:01 <kristenc> denkenz: I think what Andrew is saying is that because start and end are only 8 bit values, we can only ever have a 256 byte text described by an attribute anyway.  I went ahead and changed the array to be dynamically allocated anyway though.
Jul 01 23:15:40 <denkenz> I know what you guys are saying actually
Jul 01 23:15:46 <denkenz> So I hope my emails explained it
Jul 01 23:15:59 <denkenz> If not, let me know and we can talk here
Jul 01 23:22:08 <kristenc> denkenz: your email was not clear to me.  I'm having a hard time understanding how you would massage the start/end fields to address > 256 bytes.
Jul 01 23:22:45 <denkenz> you have to know a little how SMS works, but basically each SMS is 0..160 characters
Jul 01 23:22:51 <denkenz> Sometimes less if you have headers
Jul 01 23:23:09 <denkenz> For EMS, we need to iterate over all the headers
Jul 01 23:23:14 <kristenc> are you wanting this function to have knowledge of sms?  I thought it was meant to be generic.
Jul 01 23:23:15 <denkenz> To collect up the text attributes
Jul 01 23:23:25 <denkenz> hold on a sec ;)
Jul 01 23:23:51 * kristenc waits for enlightenment
Jul 01 23:24:15 <denkenz> Once we have collected the EMS headers, we know how much data is occupied by the actual text data
Jul 01 23:24:38 <denkenz> We thus know how many characters there are in the SMS
Jul 01 23:25:20 <denkenz> So then its a simple task of keeping the counter updated over the fragments and generate > 256 based indexes for text attributes
Jul 01 23:25:55 <kristenc> indexes for text attributes?  
Jul 01 23:26:14 <denkenz> as in you get 50 chars in first sms, 60 chars in second sms
Jul 01 23:26:29 <denkenz> we can massage text attributes in second sms to always add 50 to the start index
Jul 01 23:26:52 <kristenc> but you pass in 110 char text with all the attributes coalesced into one text attribute byte array?
Jul 01 23:27:01 <denkenz> correct
Jul 01 23:27:09 <denkenz> That is my thinking right now
Jul 01 23:27:12 <kristenc> what happens when you have a text len > 256.
Jul 01 23:27:29 <denkenz> You use unsigned short instead of unsigned char
Jul 01 23:27:35 <denkenz> so your indexes are 16 bit, not 8
Jul 01 23:27:40 <kristenc> you change the definition of a text attribute?
Jul 01 23:27:59 <denkenz> Text attribute is a data structure on the wire
Jul 01 23:28:04 <denkenz> How we implement it is up to us
Jul 01 23:29:01 <kristenc> so you are saying that instead of using a text attribute array as an argument to this function, we use our own data structure which is separate from a text attribute.
Jul 01 23:29:23 <denkenz> more or less, yes
Jul 01 23:29:34 <kristenc> since what is on the wire is going to follow the spec.
Jul 01 23:29:38 <denkenz> instead of 4 bytes, we'd need 6
Jul 01 23:29:40 <kristenc> and we are not going to.
Jul 01 23:29:53 <denkenz> err, ok, you're confusing me now
Jul 01 23:30:02 <denkenz> We're following the spec just fine
Jul 01 23:30:10 <kristenc> sorry - I am probably confused myself.
Jul 01 23:30:32 <kristenc> you said, that text attributes on the wire are separate from what we implement in our util code.
Jul 01 23:30:54 <denkenz> So something like STK Text Attribute object -> [massage function] -> our own text attributes
Jul 01 23:31:04 <kristenc> so if the we get a text attribute as defined by the spec, we can convert it into our own data struct.
Jul 01 23:31:13 <denkenz> We still implement the spec, read the wire data format, etc
Jul 01 23:31:19 <denkenz> yep
Jul 01 23:31:25 <kristenc> thus, we have no knowledge in this particular function of what the spec says.
Jul 01 23:32:02 <denkenz> Exactly, EMS would have to massage its attributes
Jul 01 23:32:10 <denkenz> and STK would as well, but the function can take care of both
Jul 01 23:32:19 <denkenz> The end result is just HTML
Jul 01 23:32:29 <kristenc> seems messy
Jul 01 23:32:55 <denkenz> so is rewriting that same function twice ;)
Jul 01 23:33:32 <kristenc> I liked andrew's idea of using a wrapper function.
Jul 01 23:33:37 <denkenz> Anyway, just an idea, don't take everything I say as canon
Jul 01 23:33:49 <denkenz> I'll accept the patch if it does 256 characters only
Jul 01 23:33:55 <denkenz> we can figure out EMS later
Jul 01 23:34:22 <kristenc> I did switch it to handle arbitrary length text array, but we still use 8 bit addressing, so effectively that did nothing.
Jul 01 23:34:41 <denkenz> exactly
Jul 01 23:35:00 <denkenz> Andrew's idea will also be messy, 'messy' cannot be avoided
Jul 01 23:35:14 <denkenz> Both will work in the end though
Jul 01 23:35:55 <kristenc> my other confusion is on the 3 vs. 4 byte attributes.  I see no way for me to know whether an attribute is 3 or 4 bytes unless it is the end attribute.  Are you saying that the caller will guarantee me a 4 byte attribute "somehow"?
Jul 01 23:36:33 <denkenz> Yeah, you're making assumptions without knowing the full context
Jul 01 23:36:50 <denkenz> In the case of EMS, you always know whether a text attribute is 3 or 4 bytes
Jul 01 23:37:04 <denkenz> Then its someone else's job to always send you quadruples
Jul 01 23:37:18 <kristenc> as I said, my understanding was that this function was to be generic and not be aware of any pdu or header information.
Jul 01 23:37:53 <kristenc> but I'm happy to assume always 4 and let someone else deal with it.
Jul 01 23:38:02 <denkenz> Just assume its always 4 for now
Jul 01 23:38:08 <denkenz> Again, lets make this work for STK
Jul 01 23:38:17 <denkenz> Go back to 256 byte static array
Jul 01 23:38:32 <denkenz> However, make sure your function is UTF8 safe
Jul 01 23:38:45 <denkenz> Since the string passed in will be > 256 bytes
Jul 01 23:40:26 <denkenz> An easy test is to use a string composed entirely of Euro signs
Jul 01 23:41:00 <denkenz> That'll result in ~380 byte string
Jul 02 01:25:01 <zhenhua> denkenz: i guess the valgrind problem is from openpty() leaks the memory.
Jul 02 01:25:50 <zhenhua> denkenz: i see bluez/tools/main.c also use cfmakeraw(). so i just grab the code from there.
Jul 02 02:10:49 <denkenz> no, it was something about uninitialized use of a structure
Jul 02 02:11:01 <denkenz> I think because you don't tcgetattr / memset the structure
Jul 02 02:17:44 <zhenhua> okay. got u
Jul 02 02:18:33 <zhenhua> for em770 patch, how about we add boolean property 'HasVoice'
Jul 02 02:19:26 <zhenhua> i mean add ENV{OFONO_HUAWEI_VOICE} in rules and add HasVoice in udev.c source
Jul 02 02:19:47 <denkenz> you mean HasVoice as modem property? That sounds fine
Jul 02 02:19:48 <zhenhua> or EnableVoice
Jul 02 02:19:57 <zhenhua> doesn't matter
Jul 02 02:45:44 <zhenhua> denkenz: questions:
Jul 02 02:45:46 <zhenhua> > +	if (g_str_equal(model, HUAWEI_MODEL_EM770W)) {
Jul 02 02:45:56 <zhenhua> > +		ofono_phonebook_create(modem, 0, "atmodem", data->pcui);
Jul 02 02:45:57 <zhenhua> > +
Jul 02 02:45:57 <zhenhua> > +		mw = ofono_message_waiting_create(modem);
Jul 02 02:45:57 <zhenhua> > +		if (mw)
Jul 02 02:45:57 <zhenhua> > +			ofono_message_waiting_register(mw);
Jul 02 02:45:57 <zhenhua> > +	}
Jul 02 02:46:18 <zhenhua> does phonebook and message waiting need voicecall capatibility?
Jul 02 02:46:36 <denkenz> phonebook no
Jul 02 02:46:46 <denkenz> message waiting doesn't make sense without voicecalls
Jul 02 02:46:54 <zhenhua> so i want to move them out
Jul 02 02:46:59 <zhenhua> okay
Jul 02 02:47:12 <zhenhua> another one is ssn
Jul 02 02:47:23 <zhenhua> it needs voicecall
Jul 02 02:47:59 <zhenhua> but we don't enable phonebook for rest huawei data only modem. do we need enable it for them or not?
Jul 02 02:50:06 <denkenz> I don't remember why we did that
Jul 02 02:50:16 <denkenz> I think it might have to do with huawei phonebook implementation being buggy
Jul 02 02:50:42 <denkenz> Try enabling it and see if it works out ok
Jul 02 02:50:57 <zhenhua> it works on em770, but not sure others
Jul 02 02:51:35 <zhenhua> yang said he tested it on em770 before
Jul 02 02:51:45 <zhenhua> so i still keep it inside if branch
Jul 02 02:51:55 <denkenz> sounds good
Jul 02 02:52:04 <zhenhua> okay
**** ENDING LOGGING AT Fri Jul 02 02:59:57 2010