IRC log of #maemo-ssu for Sunday, 2012-07-29

*** Guest38428 has quit IRC00:02
*** DocScrutinizer05 has quit IRC00:32
*** DocScrutinizer05 has joined #maemo-ssu00:33
*** NIN101 has quit IRC01:41
*** arcean_ has quit IRC04:14
*** LaoLang_cool has joined #maemo-ssu05:18
*** Guest38428 has joined #maemo-ssu05:23
*** Guest38428 has quit IRC05:23
*** amiconn_ has joined #maemo-ssu05:34
*** amiconn has quit IRC05:34
*** nox- has quit IRC05:38
*** LaoLang_coo_ has joined #maemo-ssu06:36
*** LaoLang_cool has quit IRC06:39
*** jonwil has joined #maemo-ssu06:49
*** tadzik has quit IRC07:32
*** jonwil_ has joined #maemo-ssu07:58
*** jonwil has quit IRC08:00
*** jonwil_ is now known as jonwil08:00
*** LaoLang_coo_ has quit IRC08:11
*** LaoLang_coo_ has joined #maemo-ssu08:12
LaoLang_coo_DocScrutinizer, thank you for kind help, downloading madde :)08:14
LaoLang_coo_Estel_, thank you for kind help, will try it when have time :)08:14
*** jonwil_ has joined #maemo-ssu08:16
*** jonwil has quit IRC08:17
*** jonwil_ is now known as jonwil08:18
Estel_no problem :)08:22
*** LaoLang_coo_ has quit IRC09:05
*** tadzik has joined #maemo-ssu09:44
*** LaoLang_cool has joined #maemo-ssu10:01
*** ekze has quit IRC10:16
*** ekze has joined #maemo-ssu10:16
*** Pali has joined #maemo-ssu10:29
*** LaoLang_cool has quit IRC10:55
*** jon_y_ has joined #maemo-ssu11:06
*** jon_y has quit IRC11:09
*** jonwil has quit IRC11:12
keriofreemangordon: yay for mp-fremantle-community-pr!11:35
keriojust installed, without even rebooting11:35
keriofreemangordon: why did you add the metapackage, btw?12:23
kerioevery thumb package depends on the errata kernel feature12:23
keriojust upgrading will require a good kernel12:23
*** taziff has joined #maemo-ssu12:43
*** M4rtinK has joined #maemo-ssu12:47
freemangordonkerio: the old metapackage(thumb12) was depenging on kernel-cssu = 1:2.6.28.10-cssu212:54
*** Pali has quit IRC12:58
*** Pali has joined #maemo-ssu12:58
*** M4rtinK has quit IRC12:59
*** jonwil has joined #maemo-ssu12:59
*** Pali has quit IRC13:05
keriofreemangordon: yeah but, i mean13:08
keriowhy even bother with a metapackage? won't all the packages pull an appropriate kernel via dependencies?13:08
*** NIN101 has joined #maemo-ssu13:15
*** arcean has joined #maemo-ssu13:18
*** NIN101 has quit IRC13:18
freemangordonkerio: because there is a new kernel, and if someone swithes tomorrow to cssu-thumb without having latest kernel-power installed, without having extras-devel enabled, he is busted13:21
keriohe's not, because no package from cssu-thumb will install without also installing a kernel with the errata13:21
kerioexcept for -bootimg fuckups13:22
kerioevery one of your packages depends (rightfully so) on the kernel-feature-whatever-errata13:22
freemangordonkerio: the older metapackage was depending on  kernel-cssu = 1:2.6.28.10-cssu2, the current kernel in cssu-thumb is  kernel-cssu = 1:2.6.28.10-cssu313:22
kerioi'm not talking about the particular version issue, i'm talking in general13:23
freemangordonkerio: I am not sure what you are talking about, but using -mp is the way to make system upgrades in CSSU, be it thumb or not13:25
*** NIN101 has joined #maemo-ssu13:25
keriofreemangordon: except that it worked perfectly fine for me - and i didn't have the mp- package until this morning13:25
kerioif you have cssu-testing installed, and a kernel that provides the errata, just enabling the repo and upgrading the packages that need to be upgraded will work13:26
kerioif you don't have a kernel that provides the errata, apt will want to install one, to fulfill the dependencies13:26
freemangordonkerio: STOP THAT apt-get TALKING, will you?13:27
freemangordonapt-get is not the way to make system upgrades13:27
kerio:c13:27
keriobut it has super cow powers!13:27
freemangordonAnd I don't give a ratshit how good *nix hacker you are, -mp is intended for those who are not13:30
*** rd has joined #maemo-ssu14:25
*** rd is now known as Guest1018314:25
*** Guest10183 has quit IRC14:34
freemangordonmerlin1991: ping14:38
*** M4rtinK has joined #maemo-ssu15:01
*** M4rtinK has quit IRC15:32
keriogoddammit this opname widget sucks15:34
kerioit's... wrong15:34
keriohow15:34
kerioi don't even15:34
*** jonwil_ has joined #maemo-ssu15:41
*** jonwil has quit IRC15:43
*** jonwil_ is now known as jonwil15:44
DocScrutinizer05actually I always thought MP was a Nokia invention just because they didn't like to check for proper comprehensive dependencies. And I also thought CSSU originally planned to get rid of this abomination15:45
*** asem has joined #maemo-ssu15:45
kerioDocScrutinizer05: wtf, i'm in 2g (so the only possible operator name is "TIM") but the operator name says "3 ITA CHIETI" ("CHIETI" is the cell broadcast)15:47
kerioit makes no sense15:47
DocScrutinizer05please open a ticket, please attach the logs15:49
kerioenabled logging, where are the logs saved?15:52
DocScrutinizer05(MP) I think the original problem with MP - as with any other pkg - is that there was no proper way to uninstall anything by an upgrade15:52
freemangordonqwazix, vi__: uploading fennec in thumb repo(will take maybe 15 minutes), have in mind you should manually install libqtm-1215:52
kerio"cell id change 49713" "incoming cell tower name 3 ITA CHIETI, cell id 49713"15:52
keriowhich makes no sense, because it should be "TIM CHIETI"15:52
freemangordonoops, wrong channel15:52
DocScrutinizer05kerio: complain to the BTS owner15:53
kerioDocScrutinizer05: no u15:53
kerioif i reboot i get the correct name15:53
kerioreboot/restart hildon-home15:53
DocScrutinizer05:shrug:15:54
kerioi wonder if it just caches the name because this is technically my "home network" but not quite15:55
DocScrutinizer05quite probably you're not in cell id 49713 anymore, and your current cell doesn't send any CBS on that channel you selected15:55
keriothe problem isn't the CBS, the problem is the operator *name*15:55
DocScrutinizer05please don't pester me I still got headache15:56
DocScrutinizer05the cbswidget gives no fsck about whether there's a 'name' in the broadcast15:56
DocScrutinizer05<kerio> "cell id change 49713" "incoming cell tower name 3 ITA CHIETI, cell id 49713"15:56
DocScrutinizer05I gather that's from log15:57
kerioyeah but the problem is the "3 ITA" part15:57
DocScrutinizer05so that is what you get15:57
kerioIT'S AN OPERATOR WITH NO GSM ANYWHERE15:57
kerioit can't be 3 ITA15:57
DocScrutinizer05don't shout!!!15:57
keriooh right, headache15:58
DocScrutinizer05as mentioned above, complain to BTS owner for sending BS15:58
kerio"Start Operator Name" "operator 3 ITA error 0" "oper is TIM" "code 34 240 153" "service provider name empty" "set operator name TIM"15:58
kerionotice the "operator 3 ITA" part15:59
DocScrutinizer05WE ARE NOT RESPONSIBLE FOR CONTENT OF SMSCB15:59
kerioYES YOU ARE BECAUSE YOU MADE ME INSTALL IT15:59
kerio(not you in particular)15:59
DocScrutinizer05STFU15:59
freemangordon:nod:15:59
kerio:c15:59
DocScrutinizer05disable the feature if you don't like what they send16:00
keriothe... feature? as in the operator *name*?16:00
*** asem has quit IRC16:01
DocScrutinizer05>:-(((( it's obviously not in our discretion to do any filtering to the data that gets sent by BTS16:01
keriothe bts is sending his fucking operator name correctly16:02
DocScrutinizer05except for not displaying SMSCB at all16:02
kerioi disabled smscb now16:02
DocScrutinizer05open a ticket, attach the logs!16:02
DocScrutinizer05so we can close it as invalid16:02
kerioi'm not sure you understood the problem correctly16:03
keriooperator-name-cbs-widget displays operator name incorrectly for me16:03
keriooperator *name*16:03
keriothe part that comes before any cbsms16:03
keriothe stock one displays it correctly16:03
DocScrutinizer05open a ticket, attach the logs!16:05
kerioso you can close it as invalid?16:05
DocScrutinizer05then we might investigate16:05
*** jonwil_ has joined #maemo-ssu16:07
kerioforget it, i just uninstalled it16:07
DocScrutinizer05my suspicion is that cbsmswidget doesn't properly update the name when roaming from a cell with cellbroadcast to a cell without cellbroadcast16:07
kerioyep, stock one confirmed working properly16:07
DocScrutinizer05kerio: fine, then please never again pester me with any "bug report"16:07
*** jonwil has quit IRC16:08
*** jonwil_ is now known as jonwil16:08
DocScrutinizer05as otherwise my answer will be: "forget it! just delete it!"16:08
jonwilwhats this about operator-name-cbs-widget being broken?16:09
DocScrutinizer05this is TESTING, you know? It's meant for YOUR testing and reporting about problems16:09
keriojonwil: it displays the wrong operator name for me, sometimes, when i change from 2g only to 3g only (it causes a change in the operator name because my home operator is 3g-only, and uses another 2g operator for national roaming)16:10
DocScrutinizer05if you're not willing to do proper reports, you shouldn't use T first instance. And if you do pleas edon't pester us with unproductive noise16:10
jonwilIf you wont file a proper bug, dont complain here16:11
DocScrutinizer05exactly16:11
DocScrutinizer05well, it was kinda ok'ish until [2012-07-29 15:07:07] <kerio> forget it, i just uninstalled it16:12
kerio<DocScrutinizer05> open a ticket, attach the logs! <DocScrutinizer05> so we can close it as invalid16:12
kerioit's not exactly encouraging16:13
DocScrutinizer05kerio: noooooo dude!16:14
DocScrutinizer05[2012-07-29 15:05:16] <kerio> so you can close it as invalid?16:14
DocScrutinizer05[2012-07-29 15:05:17] <DocScrutinizer05> then we might investigate16:14
DocScrutinizer05[2012-07-29 15:07:07] <kerio> forget it, i just uninstalled it16:14
DocScrutinizer05[16:14
DocScrutinizer05is the whole story16:14
keriothat's from a later point16:14
freemangordonkerio: the way you demand support is not how it should be. if you have CSSU-testing and have a problem, there is a bugtracker. period.16:15
DocScrutinizer05get lost!16:15
jonwilas for actually fixing that problem, unless either someone with good skills shows up and reverse engineers the stock widget (to figure out the exact algorithm used and find out why my algorithm is incorrect) or we can somehow get the logic from Nokia or something, I dont see hoe it can be made better16:15
keriojonwil: DocScrutinizer said something about not updating the name correctly when moving from a cell with cb to a cell without it16:16
kerioand/or viceversa16:16
freemangordonjonwil: bugreport with logs attached could help ;)16:16
jonwilyes it would help16:16
kerioI'M WORKING ON IT DAMMIT16:16
jonwilok :)16:16
freemangordonshhht, do not shout16:16
DocScrutinizer05we need to reset SMSCB display on (cell-handover + timeout-expired)16:17
DocScrutinizer05jonwil: ^^^16:17
DocScrutinizer05jonwil: where timeout is ~90s16:17
jonwilI have no way to test such a change16:17
jonwilso I cant write the code16:17
DocScrutinizer05or maybe 18016:17
keriocomponent?16:18
keriothere's no operator-name-cbs-widget16:18
freemangordongeneral16:18
kerioi'll put it in general,y eah16:18
freemangordonCSSU16:18
DocScrutinizer05jonwil: you don't see a change in CID?16:18
jonwilI do see a change in CID16:18
jonwilmy operator has stopped sending CBSMS messages for any cells that are anywhere near me16:18
DocScrutinizer05>>cell id 49713<<16:18
jonwilso I cant verify the case of "hey, moved to new cell and got CBSMS message16:19
DocScrutinizer05no, but you can verify the case of "moved to new BTS and got *no* SMSCB during 90s guardtime"16:20
DocScrutinizer05and reset to normal display in that case16:20
DocScrutinizer05I'm more than willing to do the final tests for you, I still got proper SMSCB on chan221 here16:20
jonwilThe code will reset the CBSMS if it sees a change of operator and a change of mode (2G vs 3G)16:21
DocScrutinizer05this needs to add CID16:21
DocScrutinizer05test for CID16:21
DocScrutinizer05test for change in CID16:21
jonwilWhere is the specific case of an operator sending CBSMS on only certain cells16:21
jonwili.e. who is seeing the problem in question16:22
DocScrutinizer05where kerio lives16:22
jonwilHow do we know for sure that no CBSMS is being sent? Do we have logs to confirm that yes, we see a change of cell ID but no new CBSMS?16:25
DocScrutinizer05also it's a question of design spec. Since SMSCB is BTS-specific info, it MUST get reset when handover to another BTS aka cell happens. For user convenience/experience we will want to delay that reset on cell handover by a guardtime, to allow new BTS to send SMSCB so there will be no "gaps" in display of SMSCB content16:25
jonwilMy old Motorola did not reset cell tower name on handover16:26
jonwilits a pitty there is very little documentation out there in the public space about CBSMS and its workings16:27
DocScrutinizer05jonwil:  on *any* cell handover you start a guartime timer (90s). If the timer expires you simply reset the display. When a new SMSCB comes in you delete the timer16:27
jonwilas I said, my old motorola didn't do it. But then again, it may not have been doing the right thing w/r/t the specs16:27
jonwil:)16:28
DocScrutinizer05or the guardtime been longer16:28
DocScrutinizer05actually I guess any crappy old motorola probably will do an implicit reset when T3212 expires16:29
kerioisn't there a way to detect an operator change? CB data at that point should be considered invalid anyway16:29
jonwilwe do16:29
jonwilwe throw away CBSMS on operator change16:29
jonwilso therefore if you are seeing an operator change but not seeing the correct change in either operator name or CBSMS, its a bug16:29
kerioshould i paste the logs in the description or should i attach them?16:35
jonwilattach them I guess16:36
kerioi can only attach one16:36
jonwilalso, is this the only issue with operator-name-cbs-widget or are there others?16:36
keriothe only one i had, at least16:36
jonwilok16:36
DocScrutinizer05you can attach more than one attachment16:38
DocScrutinizer05or you tar them up16:38
keriohttps://bugs.maemo.org/show_bug.cgi?id=1264516:38
povbot_Bug 12645: operator-name-cbs-widget does not display the operator name correctly16:38
jonwilok, in those logs I see nothing wrong with the cbsms as sent, just with the operator name.16:44
jonwilor is there some other different issue with the CBSMS?16:44
kerioi even disabled the CBSMS (it only works with TIM, which is not my main operator anyway, and only the channel 50)16:44
jonwilwell the cbsms doesnt matter for the operator name16:44
keriono, enabling CBSMS only leads to the correct CB message appended to the wrong operator name sometimes16:44
kerioyeah16:45
jonwilIf you are connected to 2G only and to TIM and reboot the phone (or kill hildon-home) what do you see as the operator name?16:45
kerioi see TIM16:46
keriobut then, as i change back to 3G, the name gets stuck on TIM16:47
DocScrutinizer05see? now that's a proper ticket where we actually understand what you're doing, what's expected outcome and how we could reproduce the problem16:47
jonwilkerio, I think this might be fixable if you are willing to reinstall the widget and run some specific tests.16:47
kerioyeah, i reinstalled it16:48
jonwilok16:48
jonwilstep 1, connect to 3G16:48
jonwildelete cbsms.log and opername.log16:48
jonwilor clear them out16:48
keriodone16:48
DocScrutinizer05>cbsms.log16:48
jonwilthen restart the phone16:48
DocScrutinizer05is the command ;-)16:48
jonwilor restart hildon-home16:48
kerioDocScrutinizer05: > as a shorthand for touch is neat16:49
kerioer16:49
keriofor rm+touch16:49
keriojonwil: i think that those two lead to different results16:49
jonwilyeah restart the phone16:49
jonwilthats the right answer16:49
jonwilso delete the log files and restart the phone16:49
keriobootin'16:50
keriok, booted, operator name is displayed correctly16:50
jonwilso it prints "3 ITA"?16:50
kerioyeah16:50
jonwilnow mark in the opername.log and note "booted on 3G, name is 3 ITA"16:51
DocScrutinizer05and you're on what? 2G or 3G?16:51
kerioi'm on 3G16:51
keriook, done16:51
jonwilforget cbsms.log, thats not relavent to operator name16:51
jonwilnow swich to 2G16:51
jonwiland make a note in opername.log to say that you switched to 2G and what operator name you get16:52
jonwilthen after that, restart the phone16:52
DocScrutinizer05echo "note 2G" >>cbsms.log16:52
DocScrutinizer05echo "note 2G" >>opername.log16:52
DocScrutinizer05even16:52
jonwiland make another note about what operator name you see16:52
keriohm, i see "3 ITA"16:53
jonwilok, so you are on 2G and you see 3 ITA16:53
jonwilok, make a note of that in the log file16:53
keriosometimes it works correctly16:53
jonwilok16:53
jonwilok, so what you done so far is booted in 3G with empty log file16:54
DocScrutinizer05echo "hm, i see 3 ITA" >>opername.log16:54
jonwilthen made a note16:54
jonwilthen switched to 2G16:54
jonwiland made another note16:54
kerioyep16:54
jonwilnow reboot the phone whilst still in 2G16:54
DocScrutinizer05echo "rebooting" >>opername.log16:55
jonwilmake another note that you rebooted in 2G and what operator you see16:55
jonwilafter that, switch back to 3G and make one final note16:55
jonwilthen give me opername.log16:55
DocScrutinizer05attach to ticket16:55
jonwilyeah16:56
DocScrutinizer05prepare for doing whole sequence again, in case we missed something16:56
jonwilthat sequence "boot in 3G, switch to 2G, boot again in 2G, switch to 3G" should be enough16:57
keriomy poor battery :(16:57
DocScrutinizer05jonwil: you know about the devil and where it hides. one time echo foo >bar instead echo foo >>bar is enough to spoil things16:58
keriohm, something changed16:58
kerioi get the correct name now16:58
DocScrutinizer05even better!!16:59
DocScrutinizer05so we got a proper commented log of a case where it worked16:59
jonwilyeah get me the log16:59
DocScrutinizer05of same device user location where it failed before16:59
keriothe change i see is that it says "oper is (null)" "oper is changed to TIM" ... "oper is (null)" "oper is changed to 3 ITA"16:59
keriohttps://bugs.maemo.org/attachment.cgi?id=346117:01
*** DocScrutinizer05 has quit IRC17:03
*** DocScrutinizer05 has joined #maemo-ssu17:03
kerioanyway, the operator name changes correctly now17:03
*** DocScrutinizer has quit IRC17:03
*** DocScrutinizer has joined #maemo-ssu17:03
keriosometimes it does that17:03
DocScrutinizer05your phone used different 2G BTS aiui17:09
keriohm, really? i noticed some changes in cell in cbsms.log, but they were on the 3G side17:10
DocScrutinizer0534 2 1;  34 146 1    vs    34 130 117:10
kerioi see17:11
kerioi'm in a place with not so good coverage, sadly17:11
jonwilthose numbers have nothing to do with BTS17:11
jonwilor cell ID17:11
DocScrutinizer05sad for you, good for us ;-)17:11
DocScrutinizer05ohh, then never mind17:11
jonwilI cant see for the life of me what my code is doing differently to the stock (and working) code17:14
DocScrutinizer05kerio: since the opername.log doesn't contain such interesting info like CID, could you attach cbsms.log as well? (maybe next time ;-D)17:15
keriok17:15
jonwilthe cid doesn't matter17:15
DocScrutinizer05sure it does matter, since I see if the cell changed17:16
jonwilchanges in the cell ID have no effect on the operator name selection logic17:16
DocScrutinizer05now THAT doesn't matter, since the cells themselves might behave different17:16
DocScrutinizer05and we'll have a hard time to spot this when we don't even know which CID17:17
DocScrutinizer05there's also quite some hidden 'logic' in modem itself. E.G. it stores last cell parameters to speed up a reboot and re-login17:18
DocScrutinizer05that's why your phone is 'online' way faster when switching down and restarting in same location, than when switching down and moving powered down phone several miles before switching on again17:19
DocScrutinizer05quite possible that modem behaves different regarding opername as well, when booting to same cell, than when booting to a "new" environment17:21
DocScrutinizer05it might safely assume that opername didn't change when CID didn't change17:21
DocScrutinizer05(well, CID and CC and Oper)17:22
DocScrutinizer05and last not least any such logic in modem might in fact be buggy17:23
jonwilwhat I wouldn't give for source code to connui-cellular right now (or to pick the brains of Mikko Mehtonen who wrote connui-cellular). But since that isn't going to happen the only way to solve this is to further reverse engineer libconnui-cellular or the stock widget.17:24
keriowhere does this Mikko live?17:25
jonwilno idea17:25
jonwilhe is a Nokia guy :P17:25
jonwilor was at the time :P17:25
jonwiland I have used all my reverse engineering skills already so I have no further ability to find out whats wrong with this widget17:26
jonwili.e. why my algorithm doesn't match the Nokia algortihm 100%17:27
DocScrutinizer05then maybe find out what's not working, no matter how Nokia did it?17:32
jonwilI would but I dont know whats working because I dont know how its supposed to do it17:33
keriowhere does the operator name come from?17:33
jonwilit comes from various possible sources17:33
DocScrutinizer05I mean c'mon we got a testcase as comfortable as rarely ever, with kerio producing working and not working scenarios17:34
DocScrutinizer05can't be that difficult to add proper debug printf to log, to find out WHAT differs between the working and the non-working ones17:35
kerioDocScrutinizer05: well, it would've been more comfortable for jonwil to actually experience the problem17:35
kerio:)17:35
DocScrutinizer05well, alas neither he nor me can reproduce your situation17:36
DocScrutinizer05since at least here there's no such setup where 3G differs from 2G17:36
jonwilAdding more logging wont help trust me17:38
DocScrutinizer05but he could throw in a few dozen additional debug printf17:38
DocScrutinizer05err, why?17:38
jonwilbecause there is logging in every place related to operator name that would matter17:38
jonwiltrust me17:38
DocScrutinizer05what else will help?17:38
DocScrutinizer05obviously there isn't as you can't even tell why your code delivers diferent results17:39
DocScrutinizer05or can you?17:39
DocScrutinizer05I elaborated above that for example CID might matter pretty much17:40
DocScrutinizer05since I suppose you got no random function in your code, any of the input parameters must differ to cause differing results17:41
DocScrutinizer05only way to find out is to trace as many of those input parameters as possible17:42
DocScrutinizer05maybe we can neglect sub-second timing issues17:42
DocScrutinizer05maybe we find out we actually can't17:43
jonwilThe main problem is that there is ZERO documentation on what connui_cell_net_status_*, connui_cell_net_get_operator_name, connui_cell_sim_get_service_provider or connui_cell_sim_is_network_in_service_provider_info actually DOES17:43
DocScrutinizer05that's not causing the problem, that's causing your headache17:43
jonwilThats whats causing the problem to fix this17:43
jonwilMore specifically I have no clue how the code in libconnui-cell fills in the operator_name field of the net_status callback17:44
DocScrutinizer05even kerio (no offense, kerio!) noticed a difference in logs between working and not working case17:45
jonwilyes I know there is a difference between the logs17:45
jonwilI know which code paths its following through the code in all the different cases17:45
keriooffense taken! :c17:46
DocScrutinizer05so you say you call an undocumented function() that returns opername, and this return value differs despite you feeding the function with identical parameters?17:47
DocScrutinizer05or did I get it wrong?17:47
jonwilno17:47
jonwilThe difference has to do with how the code fills in the operator_name field in the state structure passed to the widget_net_status_cb17:47
DocScrutinizer05sorry but that doesn't tell me much17:48
jonwilI mean how the code in libconnui-cell fills that in17:49
DocScrutinizer05as I haven't looked into the code yet17:49
jonwilsome code we dont have  and know basically nothing about is filling in a structure differently when it passes it to the operator widget in different cases17:49
DocScrutinizer05so how's that any different from my assumption >>you call an undocumented function() that returns opername, and this return value differs despite you feeding the function with identical parameters?<<17:50
DocScrutinizer05>some code we dont have  and know basically nothing about< == >you call an undocumented function()<;  >is filling in a structure differently when it passes it to the operator widget< == >this return value differs despite you feeding the function with identical parameters<17:53
jonwilLet me run similar tests to what kerio ran on my own phone to see what output I get in case the output I get means loading the stock widget into GDB will help17:53
jonwilok, I think I can do some gdb debugging to see whats going on here I hope17:55
DocScrutinizer05you're still too focused on diffs between your  code and Nokia code, when you rather should analyse what's exactly going on in your own code17:55
jonwilI know what my code is doing17:55
jonwilexactly what its doing17:55
jonwiland knowing that doesn't help17:56
jonwilbecause I know how its deciding which operator name to use17:56
DocScrutinizer05so which cannel is it the differences sneak in?17:56
jonwilwhat I need to figure out is under which circumstances to use which of the 3 or so possible operator names17:56
DocScrutinizer05the differences between your code working and your code not working17:57
jonwilwhich is why I am about to GDB the stock widget to find out17:57
*** arcean has quit IRC17:57
*** arcean has joined #maemo-ssu17:58
DocScrutinizer05you should focus on the differences between runs of your own code that yield different results, rather than on differences between your code and Nokia's17:59
jonwilthe differences are because Nokia functions that we know almost nothing about return different values. The stock widget clearly handles these different results in a way mine doesn't. Finding out what the Nokia widget does differently is the only way I am going to fix this.18:01
DocScrutinizer05if you can't tell the difference, you are missing some channels for input parameters18:01
DocScrutinizer05that's why I suggested to log CID and other possibly relevant parameters, as it's obvious that not all input channels are logged18:02
DocScrutinizer05and you can bet on finding in the end that Nokia's code *does* monitor those input channels you don't, as otherwise Nokia had no way to do "the right ting" where your code fails to do so18:03
DocScrutinizer05since we have logs of both working and failing scenarios, we could spot the channel that delivers differing info18:05
DocScrutinizer05at least if we'd bother to add logprints for allegedly unrelated stuff that in the end will turn out to be related nevertheless, like CID18:06
jonwilI can say the cell ID doesn't matter as none of the inputs the Nokia code receives include the cell ID (i.e. its not passed in to the callback nor does the code make any calls to functions that return it)18:08
jonwilnor is it making any dbus calls to obtain it either18:08
DocScrutinizer05of course you *might* be able to spot the missing input channel by careful analysis of that closed source Nokia blobs18:08
DocScrutinizer05jonwil: but I hope you concur that in the end it's modt likely a difference in BTS behavior that casues the problem (or at least there's a certain chance it's related), so while CID maybe isn't the missing input channel, it still may verify or falsify this hypothesis when we see one CID it always works while other CID it never works. So from this finding we might go on trace raw data delivered by cell, to find what's the difference18:12
DocScrutinizer05in format or content of the opername relevant bits18:12
DocScrutinizer05or we find it always works when BTS is same as before boot, while it always fails when BTS is a new one18:13
DocScrutinizer05again we'd like to know WTF *is* the CID so we could tell apart differing BTS18:13
DocScrutinizer05as a general rule of thumb you'll hardly learn new stuff by monitoring only those info bits you already know about18:15
jonwilClearly my code is doing something different to the Nokia code in the "it displays wrong operator name" case. Therefore (since I can get something similar to that case locally) using gdb to find out whats different makes sense18:25
jonwilI have a lead on that, I just need to get my widget into gdb to confirm something18:28
*** Estel_ has quit IRC18:29
DocScrutinizer05some background: there's a opercode->operator-name table in modem (most likely), another one in userland of AP iirc, and there's two flavours of over-the-air message with operator name (long and short operator name)18:34
*** M4rtinK has joined #maemo-ssu18:34
DocScrutinizer05whenever the values in those tables differ from what the network sends, you can see the operator name change in display a short while after login18:35
DocScrutinizer05aaaah I forgot there's also a operator name on SIM18:36
jonwilwell there are 3 different operator names we are dealing with here, one comes from what gets passed into the callback, second is what we get when we call connui_cell_net_get_operator_name and third is what we see when we call connui_cell_sim_get_service_provider18:36
DocScrutinizer05my SIM has operator name "interkom" while some tables have "o2" and other source has "O2_de"18:37
*** Estel_ has joined #maemo-ssu18:38
DocScrutinizer05connui_cell_net_get_operator_name  and   connui_cell_sim_get_service_provider seem quite obviously matching to my above info regarding name from SIM and name from network18:39
DocScrutinizer05where "from network" actually are three distinct sources OTA: operator-id numeric, operator-name-short ascii and operator-name-long ascii18:41
DocScrutinizer05form operator-id numeric both the modem and the linux may find an ascii name in local databases18:42
DocScrutinizer05s/form/from/.18:42
DocScrutinizer05in the end this is what you'll find when analyzing the Nokia opername blobs18:43
DocScrutinizer05nfc how Nokia is mixing and priorizing all those methods, and what they give back via function return value in the end18:44
DocScrutinizer05but I know on some phones I see 3 opernames in sequence: 'interkom' (from SIM), 'O2', and finally 'o2_de'18:45
DocScrutinizer05all this happens during ~20s after typing in PIN18:46
jonwilaha, I think I have a fix18:52
jonwilLet me build a new deb for kerio to test18:52
keriojonwil: there's a problem though18:53
jonwilwhat problem?18:53
kerioit's currently working fine18:53
jonwilok18:53
jonwilin that case I should just get this change into git (and hence from there into the next cssu-testing release) and then people can report any problems with this new version18:54
kerioi thought that cbs widget was going to be removed in a future version18:54
jonwilwhere was that said?18:55
jonwiland what reason to remove it?18:55
DocScrutinizer05it conflicts with custom-opername-widget, so it will not get removed but go optional18:56
jonwilI can make it support a custom name easily enough18:56
DocScrutinizer05that's been one of the suggested improvements/fixes to the issue18:57
jonwili.e. if you give it a custom name, it overrides whatever operator name it would otherwise pull from the network18:57
jonwilany other changes that would need to be made to the widget?18:57
*** Estel_ has quit IRC18:57
*** Estel_ has joined #maemo-ssu18:57
jonwilor is that the only one you know of that would need to be made for people to be happy?18:57
DocScrutinizer05atm I don't know of other major concerns18:58
jonwilok, well I can fix that easily enough with a bit of work18:58
DocScrutinizer05except of course the allegedly missing 100% functional compatibility to stock as rised by e.g. kerio18:58
jonwilalso btw DocScritinizer05, there is already logic in there to clear out the old cell tower name if there is a change of cell tower and no new CBSMS comes in after a timeout18:58
jonwilI cant remember if it was me or pali that added it but its there18:59
DocScrutinizer05please check out custom-operator-widget to see how it works, so any extensions to that one will work with yours as well (I heard of people [ab]using custom-opername-wdgt to display all sorts of weird stuff, like date, or RSS channels, or whatever. Seems there's a sort of "API" to change the displayed string at runtime)19:00
DocScrutinizer05I think it's been pali after I asked him to do that19:01
DocScrutinizer05jonwil: *I* however had a few maybe weird ideas a few hours ago: add parameters to the string to display (if you get such "constant" user defined string instead of cellbroadcast), sth like ID = cell id, B = SMSCB string as usual (but embedded into the constan string), other stuff that *might* be interesting for some users, like BCH = SMSCB channel as set in config, all the opernames, whatnot else19:06
DocScrutinizer05dafaq19:06
DocScrutinizer05jonwil: *I* however had a few maybe weird ideas a few hours ago: add parameters to the string to display (if you get such "constant" user defined string instead of cellbroadcast), sth like %CID = cell id, %CB = SMSCB string as usual (but embedded into the constan string), other stuff that *might* be interesting for some users, like %CBCH = SMSCB channel as set in config, all the opernames, whatnot else19:07
jonwilLets just get it doing custom operator names for now19:07
jonwilReading the source of custom-operator-widget, the only place it gets the string to display from is from gconf19:08
jonwiltherefore anyone changing it dynamically must be setting it in gconf19:08
DocScrutinizer05custom name: "I listened to channel %CBCH on cell %CID and got \"%CB\" as message"19:08
DocScrutinizer05jonwil: (gconf) sounds about right19:09
jonwilall that fancy stuff is way beyond what I ever intended operator-name-cbs-widget o do19:09
jonwilto do19:09
jonwilthe only reason I am adding the custom operator name thing is because lots of people seem to want it :)19:10
*** andre__ has joined #maemo-ssu19:31
jonwilok, if this compiles, I can test it to make sure the functionality works, fix it if its broken, then commit it to git19:39
jonwilonce this is done, we can push this to testers in next cssu-testing and if they want more functionality, they can ask for it and it will be added19:40
*** nox- has joined #maemo-ssu19:50
freemangordonjonwil: better push it in cssu-devel19:51
freemangordonso it'd be tested before next -testing19:51
jonwilwell it will be going into the relavent git19:53
jonwilhttp://gitorious.org/community-ssu/operator-name-cbs-widget19:53
jonwilwhatever people do with it from there is up to them19:53
freemangordonjonwil: cssu-devel is not a git repo, but debian repo ;)19:53
freemangordongit repo remains the same19:53
jonwilyes that I know19:53
freemangordonthen I am unable to parse19:54
jonwilits up to someone who knows all the repos to make sure it gets from the git repo into whatever other repos it has to go into19:54
jonwilsince these debian repos are not my responsibility19:54
freemangordonno, it is you to do it. though you can ask merlin or Pali19:54
freemangordonask merlin1991 then19:54
jonwilok, just need to find out why its not properly repainting the widget when the custom name is changed20:08
amiconn_jonwil:  I do have the problem with the cssu-provided operator name widget.20:14
jonwilok, what problem do you have?20:14
amiconn_My home provider sends no cell broadcasts at all. Normally this doesn't affect me, but two weeks ago I've been in Poland (i.e. international roaming). After returning to Germany (far enough from the border so I've  definitely been connected to my home provider) the widget didn't show my home provider name. It still displayed the Polish provider's name20:17
jonwilI assume your phone was not restarted or turned off between leaving poland and noticing this problem?20:18
amiconn_Stock widget does not have that problem20:18
jonwilWas your phone turned off or restarted after leaving poland and connecting to the german operator?20:19
amiconn_no20:19
jonwilok20:19
jonwilI have a fix that will be going into git shortly that should fix your problem if is the same one I have been seeing in logs before20:20
jonwili.e. if its the one I am thinking it is20:20
amiconn_I did this after noticing the problem. This fixed the display20:20
jonwilok, that confirms that its most likely the issue I am thinking of20:20
* amiconn_ would like to uninstall the widget, because he has no use for it, but I don't know whether that's possible without uninstalling other parts of cssu...20:21
jonwilok, that all works, now I just have to remember how to commit to git20:21
*** amiconn_ is now known as amiconn20:21
*** mase76 has joined #maemo-ssu20:35
*** M4rtinK has quit IRC20:45
jonwilok, I committed my changes to git20:59
jonwilmerlin1991: ping21:02
kerioamiconn: it's possible to remove it, the only side effect would be that it uninstalls the mp- package21:10
kerioand you'd possibly have issues if you autoremove afterwards21:10
kerioanyway, amiconn <321:11
kerioI AM NOT CRAZY!! ^_^21:11
*** andre__ has quit IRC21:32
*** jonwil has quit IRC21:39
*** arcean_ has joined #maemo-ssu21:58
DocScrutinizer05amiconn: you rather update to latest jonwil version ;-)21:59
DocScrutinizer05kerio: nobody doubts you're crazy ;-)21:59
DocScrutinizer05kerio: we knew about problems in cbsms-widget already22:00
DocScrutinizer05I went thru all this with amicon before22:00
DocScrutinizer05and pali implemented something I thought might fix the issue22:01
*** arcean has quit IRC22:01
DocScrutinizer05I just forgot I suggested that fix and pali actually implemented it, until jonwil mentioned there's already such function22:02
*** nox- has quit IRC22:13
*** KrayonWork has quit IRC22:13
*** nox- has joined #maemo-ssu22:17
*** Jade has joined #maemo-ssu22:58
*** Guest10183 has joined #maemo-ssu22:58
*** arcean_ is now known as arcean23:20
*** mase76 has quit IRC23:49

Generated by irclog2html.py 2.15.1 by Marius Gedminas - find it at mg.pov.lt!