IRC log of #maemo-ssu for Sunday, 2015-10-25

Sicelo009Nas freemangordon said, the 'emergency-ness' of the call is easily apparent on pin entry ... because at that point the phone should not let you do any calls via the SIM whose PIN you probably even do no know. but at least you can use it for life & death situations00:00
jonwilfreemangordon: where is _connui_internet_status_menu_item_is_suspended?00:01
jonwilyou missed that one00:01
freemangordonhmm, right00:02
freemangordonI wonder why -Werror didn't catch that :(00:02
freemangordonjonwil: going to add it00:03
jonwilok00:03
bencohSicelo: my point was that nowadays phones are usually turned on but password-protected ... not protected by the sim's pincode00:06
Siceloeven then .. the emergency call is (should) still be possible00:06
*** handaxe has joined #maemo-ssu00:16
freemangordonjonwil: done00:17
*** handaxe has left #maemo-ssu00:18
jonwilgreat00:18
jonwilnow we just need the debian packaging00:18
freemangordonI'll do it, probably tomorrow00:18
freemangordonstill, why -Werror doesn't catch functions with no prototypes?00:19
freemangordonwell, actually there is no warning at all00:19
*** Pali has quit IRC02:37
*** LauRoman|Alt has quit IRC03:28
*** LauRoman has quit IRC03:28
*** futpib has quit IRC03:39
*** sparetire_ has quit IRC05:41
jonwilI just cloned osso-systemui-modechange :)06:38
Sicelo:)07:19
*** Sicelo009N has quit IRC08:24
*** Sicelo009N has joined #maemo-ssu08:36
*** Sicelo009N has quit IRC08:46
freemangordonjonwil: why is that? https://github.com/community-ssu/osso-systemui-alarm/commit/ec77165f915698c96aede6796d8cdb1d4fb4253108:50
freemangordonjonwil: those GtkWidget->GtkWindow are simply wrong, please revert those commits and use GTK_WINDOW(obj) cast when needed08:57
freemangordons/are simply/changes are simply/08:57
infobotfreemangordon meant: jonwil: those GtkWidget->GtkWindow changes are simply wrong, please revert those commits and use GTK_WINDOW(obj) cast when needed08:57
freemangordonsee https://github.com/community-ssu/osso-systemui-alarm/commit/ec77165f915698c96aede6796d8cdb1d4fb42531#diff-845e439430941a41c60246e333118df9R95008:58
freemangordongtk_dialog_new() returns GtkWidget*, if you want to use that object as anything else, you should cast it with gtk macroses like GTK_WINDOW(), GTK_DIALOG() etc09:00
jonwilok, will undo the change09:00
freemangordonc-style cast is not a good idea, as GTK_XXX does a runtime checks if the object can really be converted to the required type09:00
*** Sicelo009N has joined #maemo-ssu09:00
freemangordonjonwil: just explaining why :)09:01
jonwilok09:01
jonwilok, I fixed it09:03
freemangordonjonwil: when you are sure everything works as it should, you may add "-DG_DISABLE_CAST_CHECKS" to CFLAGS, then runtime checks are not performed and GTK_XXX turns into a c-style check. see https://github.com/community-ssu/rtcom-accounts-plugin-gtalk/blob/master/configure.ac#L50 and bellow on a sane way this to be done :)09:04
freemangordongood09:04
freemangordonjonwil: btw, why did you do this change?09:05
jonwilI found clear evidence that those 4 systemui functions actually do take a GtkWindow and not a GtkWidget. More specifically,  I found code that called those functions passing in GTK_WINDOW(blah)09:06
jonwilwhich wouldn't be appropriate if it was a GtkWidget going in09:06
freemangordonjonwil: hmm, a simle "git revert" should've been better, as commit description "more fixes" doesn't tell us much :), sorry to be picky, but in a month from now, we will have no idea what and why it was done :)09:07
freemangordonjonwil: then function should be declared as f(GtkWindow *) and all the callers should cast to GtkWindow09:08
jonwilall 4 functions are declared as f(GtkWindow) now09:08
freemangordongreat09:08
jonwilthe only mistake I made was not doing the right thing in osso-systemui-alarm09:09
jonwilI got confused because osso-systemui-alarm seems to be compiled with -DG_DISABLE_CAST_CHECKS09:09
jonwiland didn't realize I needed some macros in there09:10
freemangordonwhich we should use for all of the packages, but this is another story09:10
freemangordonjonwil: oh, and please, change debian/changelog files and put your name here, not mine or Pali's :p09:11
freemangordon*there09:11
jonwiloh sorry09:11
jonwilI used dch command to update changelog and didnt even notice the names09:11
freemangordonwhat is dch?09:11
freemangordonI always edit changelog files by hand09:12
* freemangordon finds gtk type macros overly confusing sometimes :)09:13
jonwilanyhow, osso-systemui-alarm is using the right things now09:14
freemangordonjonwil: on a side note, how it is the correc way (in english, for plural): macroses or macros?09:14
jonwilmodechange, tklock and powerkeymenu are fine as they didn't use G_DISABLE_CAST_CHECKS and hence I didn't get confused with either of those09:14
jonwilmacros09:14
freemangordonok, thanks09:15
freemangordonoh, yeah, it is macro for single, weird, as in Bulgarian we say "macros" for single. however09:15
freemangordonjonwil: this define swithes *runtime* checks off, not compile time, iirc09:16
freemangordonor there are warning in compile time as well?09:16
freemangordon*warnings09:17
jonwilthere are compile time warnings if you pass the wrong pointer (e.g. a GtkWidget when it wants a GtkWindow)09:17
freemangordonoh, good09:17
freemangordonjonwil: actually what this "modechange" plugin does?09:18
jonwilIf you are in airplane mode and you do an action that goes out of airplane mode (e.g. trying to activate a data connection) it displays a box saying "are you sure you want to leave airplane mode"09:19
jonwilWhich doesn't happen if you press the "exit airplane mode" button on the powerkeymenu09:19
freemangordonI see09:19
jonwileasiest way to see is to press powerkey, press "offline mode" then press status bar then press "internet connection" button09:20
freemangordongot it09:20
freemangordonjonwil: BTW, could you share you hildon-whatever-nsv-something IDA db, to see if I can RE it further09:22
jonwilhttps://drive.google.com/file/d/0B9idqO9KygGsM0FoUW1GU0RNbm8/view?usp=sharing09:24
jonwildoes anyone know what "act dead" state is?09:24
freemangordonsure09:25
freemangordon:D09:25
jonwilwhat is it?09:25
freemangordonjonwil: when you turn you device off and connect a charger09:25
freemangordonit turns on end enters "act dead" state09:25
freemangordonor when an alarm is triggered while it is turned off09:25
*** DrCode has quit IRC09:27
jonwilwhy it needs a systemui plugin (especially one that has UI) I dont understand09:27
freemangordonwho needs it? systemui?09:28
jonwilyes there is a systemui plugin for actdead09:28
freemangordonwell, take alarmui for example - it has a dialog asking you whether you want to "turn on the device"09:29
freemangordondunno what actdead systemui does, but I guess we can find that easy09:29
freemangordonjonwil: keep in mind that ACT_DEAD mode has almost everything up and running, including X, dbus, etc. only hildon-desktop is not started afaik09:31
jonwilIs there a difference between act_dead mode and the "battery is too low to run normal system" mode?09:32
jonwilthe one you get when your battery dies and your phone switches off then you plug it into a charger and it goes into a charging mode09:32
freemangordonyes, those are different09:32
freemangordonyou enter ACT_DEAD when there is enough battery to boot maemo09:32
freemangordonbut for some reason (like alarm triggered while off) you don;t want to enter USER mode09:33
jonwildoes ACT_DEAD mode turn on radios or does it act like airplane mode?09:34
freemangordonwhen your phone dies because of a flat battery and you connect a charger, it is NOLO that does the so-called precharging, until there is enough charge to boot maemo and enter ACT_DEAD09:34
freemangordonafaik radis are off09:34
freemangordon*radios09:34
freemangordonso yes, it is like airplane mode in that regard09:35
jonwilradios being off makes sense otherwise an alarm could cause your phone to switch on at a point where cellular radios are banned for some reason (e.g. on an airplane)09:35
freemangordonyeah09:35
freemangordonthat is why you're presented with a dialog asking you if you want to turn the device on09:36
freemangordonacually, if you select "no", the device is not turned off again, iirc :)09:36
*** Sicelo009N has quit IRC09:36
jonwilanyhow, my next potential targets are the remaining osso-systemui pieces (main daemon, actdead, devlock, emergency, splashscreen) and also possibly osso-applet-devicelock, libdevlock and libcodelockui09:40
jonwildevice lock UI would be useful since IIRC someone said it has problems in CSSU in portrait mode09:40
freemangordonisn't devlock == tklock?09:41
freemangordonor this is the one to enter the code?09:41
jonwilyeah devlock is the code entry09:41
jonwilfor security code09:41
jonwilthere is libdevlock which handles storing the code in CAL09:41
jonwilthere is libcodelockui that displays the UI for it09:42
jonwilplus the systemui stuff for displaying the unlock dialogs09:42
jonwiland the control panel for setting the lock code09:42
freemangordonok09:43
jonwilthe same UI widget is used for pin code lock if your SIM is protected by a PIN I believe (connui-conndlgs-cellular does that bit)09:43
jonwilNot so sure the main systemui daemon is really worth reverse engineering actually09:49
freemangordonwhy not, it was a small binary iirc09:50
freemangordonwhat the?!?09:50
freemangordonjonwil: could you look at system_initialize in ...-nsv09:50
freemangordonI think there is a memleak09:51
freemangordonoh, no09:51
freemangordonmy bad09:51
jonwilyeah systemui is a small binary but that doesn't mean its easy to RE09:52
jonwilI think the device lock stuff is a nice target to do next09:52
jonwilsince there is a use for that (portrait mode device lock fixes)09:52
freemangordonyeah09:52
*** mirda has left #maemo-ssu10:35
bencohfreemangordon: how do you display g_debug/g_warning? I guess I'm missing some magic envvar or args10:47
bencoh(with this version of gtk/glib ...)10:47
*** Pali has joined #maemo-ssu10:52
jonwilok, so I figured out that the list of emergency numbers obtained by libcodelockui is only used to determine whether to display the "emergency call" button and that pressing that button doesn't pass what you typed into the unlock code box (911/112/whatever) to the phone subsystem, it passes a special magic string urn:service:sos which dials the real emergency number (oops). So my hack to make...11:15
jonwil...it think "12345" was a valid emergency number (so I could test pressing the button and see it dial the invalid number 12345) went wrong.11:15
jonwilSo from now on I wont touch that button at all11:15
jonwilAlthough I just found out that if you e.g. use dbus-send via SSH to dial a regular non-emergency phone #, it will properly dial even with the phone in locked state11:35
jonwilthat said, if someone has ssh access to a locked phone and knows how to send a "dial number" command, they can also find a way to unlock the phone that way11:35
bencohjonwil: ooooooh11:41
bencohso that's how it works.11:41
jonwilnote that rtcom-call-ui has special UI if the phone is locked and you are dialing a non-emergency number to indicate its locked and give you an option to activate the unlock screen11:47
jonwilSo glad we have a reverse engineered MCE to play with, so many of the cool things I am fiddling with are tied into MCE11:52
*** jonwil has quit IRC13:22
*** futpib has joined #maemo-ssu13:38
*** sparetire_ has joined #maemo-ssu15:12
*** Sicelo009N has joined #maemo-ssu15:15
*** xes has quit IRC18:53
*** xes has joined #maemo-ssu19:13
freemangordonbencoh: nothing special, besides syslog19:45
freemangordonkeep in mind you have to reboot the device after installing syslog19:45
bencohhmmm ... then I wonder why I dont see g_warning/g_debug messages in a custom gtk program (running from cli)19:46
bencohoh ... how comes?19:46
freemangordonbencoh: you should use g_print, if you want those messages in the console19:46
freemangordonbencoh: device needs a reboot after installing syslog for the first time, bon't as me why :)19:47
freemangordon*don't19:47
bencohbut ... they're debug messages ... they shouldn't be displayed unless someone really wants it (?)19:48
freemangordonbencoh: usually you won;t see g_debug messages19:48
freemangordonthus you'd better use g_warning :)19:49
bencohI dont see g_warning either19:49
bencohthat's what I dont understand19:49
freemangordondid you reboot?19:49
bencohI guess I missed something :)19:49
bencohlaunching from cli19:49
freemangordonno matter, did you reboot?19:49
bencohwhat?19:49
freemangordondid you restart the device?19:50
freemangordonafter installing syslog19:50
bencoherrr ... do you mean every single g_* message goes to syslog?19:50
bencohand there is no way to override that?19:50
freemangordonthere is a way19:50
freemangordonbut yes, every g_... message goes to syslog, besides g_print()19:50
bencohokay ... thx for the info :)19:51
bencohI guess I really need syslog then19:51
bencoh(not just for this mafw debugging stuff)19:51
bencohoh and ... not related to this, but what prevents us from shipping libgcc-4.7 in cssu (not just thumb)?19:52
bencohis there some compat issue?19:52
freemangordonnothing afaik19:55
freemangordonwe just need time and will19:55
bencohokay19:57
bencohdo we have a non-thumb libgcc-4.7?19:57
freemangordonwe have only non-thumg libgccc19:57
freemangordonhmm, typo day19:58
bencoh:)19:58
bencohso we could just copy the package to -testing ?19:58
keriocan we have a thumb libgcc-4.720:05
freemangordonkerio: don't think so20:07
freemangordonor rather - don;t know how to do it20:07
kerioalso can we have a thumb xorg compiled with 4.720:12
freemangordonwhy not?20:13
freemangordonI just need to find some spare time20:13
freemangordonand that's not easy :)20:13
bencohsetting up 4.7.2-thumb toolchain here :)20:16
freemangordonbencoh: you'd better find someone that can build toolchain, having 4.9 would be good IMO20:17
bencohwhat do you mean by "can build toolchain"?20:17
kerioclang20:18
kerioclang is cool20:18
keriofuck gnu20:18
bencohkerio: errr ...not today(tm) :p20:18
bencohfreemangordon: "building" a toolchain isn't a real issue for me, but "packaging" it to make it work with sb is something else20:19
bencoh(oh, and we'd still need to apply err657417 patch to gcc)20:20
bencohunless this has been worked around in 4.9?20:20
freemangordonвхат ис err657417?20:23
freemangordonsorry20:23
freemangordonwhat is err657417?20:23
freemangordonbencoh: ^^^20:23
bencohhttp://wiki.maemo.org/Community_SSU/Thumb#ARM_errata_65741720:26
freemangordonbencoh: the errata workaround is in ld20:27
freemangordonand is there since 2.20 iirc20:28
bencohhmm20:29
bencohthen why do we need a different toolchain?20:30
freemangordonbecause 4.9 has more ARM optimizations than in 4.7. also, auto-vectoriser actually works afaik20:30
bencohI mean 4.7 vs 4.220:30
freemangordonbecause 4.2 has a couple of thumb-related bugs20:31
bencohoh okay20:31
bencohhm and looks like it shipped ld 2.1820:31
freemangordonthat one too20:31
bencohanyway ... what would we need to build a 4.9 toolchain>20:32
bencoh?20:32
freemangordonno idea, it was tricky20:32
freemangordonI tried once to build 4.7.2 as debs, but there was some problem when installed in SB20:33
merlin1991crosstools-ng and some debian packaging for the easy route ;)20:33
freemangordonmerlin1991: SB has that packaging already (or some HOWTO, can't remember)20:33
freemangordonbut setting all the build dirs, root, etc is tricky20:33
freemangordonalso, -ng don;t do it iirc, you need to do it by hand20:34
freemangordondon;t do it correctly I mean20:34
freemangordonalso, one should statically link host computer glibc etc, otherwise it won't run in SB20:34
freemangordonetc, etc,...20:35
merlin1991hm I got crosstols-ng built arm toolchain running in sb2 but then again sb2 works quite differnt to sb20:35
merlin1991never tried to load a different toolchain into our trusty scratchbox20:36
kerioi thought it was karmic, not trusty20:37
keriohuehue20:37
bencohkrkr :)20:37
Sicelofun times20:40
SiceloOct 25 20:38:41 fremantle kernel: [ 2438.567230] memory_sync_page: no page for 4395b00020:40
SiceloOct 25 20:38:41 fremantle kernel: [ 2438.567260] proc_memory_sync: InValid address parameters 4395a800 e8020:40
kerioi wonder what would be a decent way to modernize maemo20:43
merlin1991upgrade alll the libs?20:43
keriowhat do you do with the statically-compiled, closed stuff?20:43
kerioi guess you keep the old libs around20:43
bencohif they're statically-compiled you dont need the libs ;)20:44
kerioer yea20:44
bencohbut I guess you meant "linked against old libs"20:44
kerioyep20:44
kerioi doubt there's much that's statically compiled20:44
keriogiven that space was (and is) quite a premium20:44
bencohyou check ABI compat and symbol dependencies20:44
merlin1991kerio: did I mention, re all the bits?20:45
keriolet's just move to freebsd20:46
kerioi bet the nokia stuff runs fine in the linux compat layer20:47
bencohmeh :D20:47
drathirsad that maemo isnt based at at alar, but on the debian...20:57
drathiralarm*20:58
*** Sicelo009N has quit IRC21:26
*** jonwil has joined #maemo-ssu22:04
jonwilhi22:12
*** Sicelo009N has joined #maemo-ssu22:21
Palihi!23:08
jonwilTime to clone osso-systemui-devlock next I think, that and libcodelockui23:15
jonwilThen someone can add portrait device lock support to CSSU23:15
jonwilosso-applet-devicelock is on the todo as well23:16
jonwilProbably wont let you do portrait support for the other code lock stuff (i.e. sim pin codes) but who uses those anyway...23:16

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