IRC log of #harmattan for Thursday, 2011-08-25

*** piggz has joined #harmattan00:40
*** rm_you| has joined #harmattan00:48
*** neal` has joined #harmattan00:49
*** neal` is now known as neal00:49
*** CepiPerez has joined #harmattan00:53
* lardman curses not being able to see the edges of any of his QML elements00:59
lardmanand therefore having no clue whether they are actually extant01:00
rcgis there any way to detect when an app is "minime01:03
*** piggz has quit IRC01:03
rcgzed" and maximized01:03
ajalkanercg: yes, see the QmlComponentGallery apps01:03
rcgerr.. accidently hit enter.. but i guess you get the meaning ;)01:03
rcgajalkane: great, will do01:03
ajalkaneThere's the Visibility tab that shows how to do it01:03
lardmannight all01:04
*** lardman has quit IRC01:04
rcgajalkane: ah, sweet! thats exactly what i was looking for. thanks :)01:05
*** sudanix has joined #harmattan01:05
MohammadAGerr, sorry about that01:06
*** crevetor has quit IRC01:12
*** deimos_ has quit IRC01:23
npmrm_you: fyi, i uploaded a new version of (i fixed the prob. with all my .svn files getting copied into the deb and doubling the size)01:24
npmnow to tame the .spec file01:25
npmrm_you: you mentioned you added the last version to your repo, this one is better since it's half the size and i fixed missing icons, etc.01:26
*** andre__ has quit IRC01:38
*** mtd has quit IRC01:41
*** baraujo has quit IRC01:53
*** lcuk has quit IRC01:55
*** rcg has quit IRC01:57
wazdSteve Jobs quits as Apple CEO02:16
wazdthe world is safe now I guess :)02:16
*** jkt has quit IRC02:19
mikhashm, so in one week I can buy Apple shares then02:19
berndhsyeah but you never know02:19
berndhscould be factions, fragmentation, and all that02:20
berndhsno wait that was Libya, sorry02:20
*** jkt has joined #harmattan02:26
*** jkt has joined #harmattan02:26
MohammadAGhiemanshu, can you hide the status bar in irc chatter when the vkb opens in landscape mode?02:30
MohammadAGmost (all?) stock apps behave like that02:30
mikhaswell, if you use LMT or Qt Quick Components02:31
MohammadAGi'm guessing it's an mtf thing?02:31
*** smoku has left #harmattan02:32
*** seeder has joined #harmattan02:39
*** berndhs has quit IRC02:43
*** M4rtinK has quit IRC02:47
*** Guest84219 has quit IRC02:56
*** CepiPerez has quit IRC02:56
*** npm has quit IRC02:56
*** npm has joined #harmattan02:56
*** seeder has left #harmattan02:58
*** seeder has quit IRC02:58
*** npm has quit IRC03:01
*** npm has joined #harmattan03:06
*** joejoe has quit IRC03:09
*** joejoe has joined #harmattan03:09
*** NIN101 has quit IRC03:28
*** willer_ has quit IRC03:29
smokexif I were using qt designer based windows to develop a harmattan device instead of qt quick, how would I access that grey bar, add menus to it, etc03:39
*** djszapi_ has quit IRC03:43
javispedroyou cannot03:43
javispedro(well, obviously you can, but it's kinda rocket science)03:44
*** wazd has quit IRC03:44
*** djszapi_ has joined #harmattan03:47
djszapi_javispedro: I was wrong. Processes try to inherit the right by default.03:48
djszapi_* credentials, that is03:48
smokexI like rocket science :D03:49
smokexso could you give me a general idea of where I would start03:49
javispedrohiding that bar via fullscreen and rendering your own =)03:50
javispedrodjszapi_: is that documented anywhere?03:50
trx(then you loose the status bar too)03:50
djszapi_javispedro: read the code as I did.03:50
javispedroah yeah, you have to render the status bar too, but you can get it from a shared pixmap that is on some window property.03:51
*** djszapi__ has joined #harmattan03:51
javispedroI do not think you can use that bar for anything other than the back button03:51
djszapi_javispedro: it is documented internally, but I have not had access to that right away. I am not sure about the public SDK.03:51
djszapi_it does not mean I could reproduce your dmesh issue though.03:52
javispedroI wouldn't call it an "issue". But at least it is partially explained by that.03:53
djszapi_no, that is not an explain for it03:53
smokexok so fullscreen hides it nad tbh I don't need or want it :D thx03:54
javispedrotrx: to sum it up, if you use plain Qt you face the same problems someone using Gtk+ will face, so, take care.03:54
djszapi_we do not do logging for that, that is up to the application developers.03:54
smokexok thx its just a simple app to keep a transaction register03:55
djszapi_javispedro: it is actually a serious issue.03:55
djszapi_javispedro: since it should really not try to inherit any credential if the hash and other validations fail. I think it is a bug of your version03:56
javispedrodjszapi_: I was using --relaxed-exec fwiw03:56
javispedroso potentially the hast test passed03:56
djszapi_duh..why did you not tell it to me ? :)03:56
javispedroI _always_ use relaxed exec ;)03:57
djszapi_please provide precise scenario next time (including everything).03:57
djszapi_this way, it was a lot of wasted thinking :)03:58
*** djszapi__ has left #harmattan03:58
javispedrobut you discovered there's some kind of inheriting =)03:58
djszapi_that is a separate thing which reveals a bug here for us03:58
djszapi_oh no, it is fine with relaxed stuff03:59
djszapi_but yes, check out out gdb and develsh03:59
djszapi_you can always limit this inheritance for sure04:00
djszapi_and aegis-manifest-dev might even do that.04:00
djszapi_but it is also very easy to manually add a non-add policy really, default empty request block04:01
djszapi_javispedro: I think it is actually weird since I would expect non-inheritance by default04:05
djszapi_at put this logic into the manifest autogeneration as the default, but not as a default option for every binary.04:05
djszapi_it was not probably a high priority task to do it that way.04:09
javispedroit is probably easier for some people who forgoet to put each binary in the manifest, and probably harder for those who have many privileges and spawn lesser-privileged binaries =)04:09
djszapi_since it is not nice in theory, but since you have quite a few protections for parents as well, it is not a simple thing to think of a use case scenario04:09
djszapi_javispedro: it is not easier at all04:09
djszapi_javispedro: if you use the Qt SDK, it uses the autogeneration anyways.04:10
djszapi_so this logic would transparent, you would not forget it. Maybe you can forget to put aegis-manifest-dev into the control file if you are using scratchbox, but closing a security hole is more important than that.04:11
djszapi_and it is really transparent by using the Qt SDK.04:11
djszapi_and if you have a request blog for other reason without add policy, you do not even need the aegis-manifest-dev.04:11
djszapi_block* :)04:13
djszapi_mmm, maybe easier to port applications this way.04:16
djszapi_and since the shells for instance from known source origin are confined enough regarding the credentials, and shells and other processes from unknown origin cannot be granted with every credentials, it makes the priority lower, and a use scenario harder.04:19
*** leinir has quit IRC04:21
MohammadAGyou know what'd be awesome when multitouch is used? manipulating two things at once04:35
MohammadAGlike, suppose i'm swiping, while i'm doing that, another finger can be scrolling in the window that's being swiped04:35
javispedrothat leaves you with 8 free fingers!04:37
MohammadAGthe task switcher can have those :p04:38
GAN900MohammadAG, there's a certain point where multitasking becomes a sickness04:38
GAN900But, yes, I still wish the N800 had had two dpads.04:38
javispedroalso ,you can use the n950 with one hand and the n900 with the other hand04:39
MohammadAGGAN900, i think i hit that point when I started carrying two phones to get the job done04:39
GAN900MohammadAG, I want retail firmware. :<04:39
MohammadAGjavispedro, still haven't figured out how to use them with the N8604:39
GAN900One of my friends used to play emulators on the bus with a Wiimote in one hand and an N800 in the other.04:40
javispedrohey, I think I know that guy.04:41
MohammadAGthat's something i wouldn't do in my own car04:41
javispedroor a similar person who emailed me about a similar usecase.04:41
MohammadAGone question04:41
MohammadAGhas anyone figured out wtf the blinking blue internet icon means04:42
javispedroit means: someone is fscking with your wi-fi.04:42
javispedroI usually get it when SNR gets low and soon it'll disconnect04:42
javispedroactually I think it already signals it's disconnecting.04:43
javispedronext time check /var/log/syslog :)04:43
MohammadAGbut i'm connected to wifi and the 3G icon blinks04:43
* javispedro ponders how to map the compass usefully as a joystick device04:43
* MohammadAG wants nice compass graphics04:44
MohammadAGQMLCompass: one of the apps that should be portrait only but isn't04:45
javispedroomg this is gonna be kickass.04:48
javispedrowait, nothing.04:48
mikhasMohammadAG, if only we hadn't deactivated multitouch in most of the core UI …05:19
mikhasduring transitions, touch events are usually eaten up by the scene05:20
*** Evgeniy has joined #harmattan05:27
*** meegomy_ has joined #harmattan05:32
*** berndhs has joined #harmattan05:48
javispedrook, seriously???05:50
javispedrothe logic that converts raw magnetometer values into bearing is PROPIETARY???05:51
javispedroas is the one that gives out the supposedly calibrated magnetometer readings05:54
* javispedro had like 17 calculator windows opened, clearly a sign of too much hacking for today.06:06
*** twoboxen has joined #harmattan06:08
meegomy_have anyone involve in meego meetup?06:12
rm_you|GAN900: yo06:20
rm_you|got kicked off earlier06:20
rm_you|what did you say about SnapGo? :P06:20
*** sudanix has quit IRC06:23
*** javispedro has quit IRC06:27
*** twoboxen has quit IRC06:30
*** ajalkane_N950 has quit IRC06:50
hiemanshuMohammadAG: I'll look into that07:00
*** mikhas_ has joined #harmattan07:06
*** mikhas has quit IRC07:09
specialis Camera using qtmobility to access the camera?07:11
*** berndhs has quit IRC07:13
*** berndhs has joined #harmattan07:22
*** DocScrutinizer has quit IRC07:30
*** DocScrutinizer has joined #harmattan07:31
*** hardaker has quit IRC07:33
*** twoboxen has joined #harmattan07:33
*** twoboxen has quit IRC07:35
*** berndhs has quit IRC07:38
*** harbaum has joined #harmattan08:16
*** lardman has joined #harmattan08:19
*** lardman has joined #harmattan08:19
ajalkanehow do you do08:22
*** harbaum has quit IRC08:23
lardmanhi rm_you08:24
lardmanajalkane: not bad, you?08:24
ajalkanea bit tired, but otherwise excited to be bored at work08:25
* lardman is also rather tired, but in training for a new born who he hopes will soon arrive08:26
*** maxw has joined #harmattan08:27
Ronksu"excited to be bored at work" sounds a bit mondayish :)08:27
ajalkaneExciting times ahead then. Boy or girl?08:28
lardmanI'm having troubles remembering what day of the week it is having been on holiday for 2 weeks now08:28
ajalkaneYeah, lately all work days have felt worryingly mondayish08:30
MohammadAGno sleep for a day, fun times08:31
lardmanQML is taunting me with cool looking UI effects and all I can see is a blank screen08:31
MohammadAGrun the app with -style cleanlooks08:33
ajalkaneGood old blank screen. At least it usually clearly points to the problem position08:33
MohammadAGwithout that i don't get errors08:33
lardmanMohammadAG: on device?08:33
MohammadAGyeah, i have meegotouch-qt-style installed08:33
* lardman removes all the ui components and goes back to coloured Rectangles08:34
MohammadAGno no no08:34
MohammadAGit should say which line the problem's in08:34
lardmanthere is no problem08:34
ajalkaneListView taunts me with its inability to specify cool move/remove effects08:34
lardmanjust doesn't show anything on one of my pages08:34
MohammadAGno javascript logic either i guess?08:34
lardmanListView is one of the things I have on my page08:34
*** spenap has joined #harmattan08:34
MohammadAGif only there was ListWidget08:35
MohammadAGi hate making models08:35
ajalkaneyeah models are a bore when anything dynamic is required08:36
lardmanmore of a pita when you don't need anything very complex08:36
lardmanthe docs could do with some better intro into how to use the model view stuff imo08:36
MohammadAGwell, i could do all lists with a QListWidget in C++08:37
MohammadAGQListWidgetItem 'item = new QListWidgetItem(), set all the stuff i need, and i'm done08:37
MohammadAGthat's usually 3 to 4 lines of code08:37
lardmanDelegates are simpler in QML at least08:38
lardmanhmm, this is quite annoying08:40
ajalkaneListView can eat JS arrays as model if I remember correctly? That would do for simple case08:40
lardmanI have Pages {} in seperate qml files and I instantiate them in a TabGroup - this works if I create Pages inline, but not if I try to create from the files08:41
*** piggz has joined #harmattan08:41
lardmanam I right in thinking that I ought to be able to generate a component housed in a SummaryWindow.qml file by sticking SummaryWindow {id: summaryWindow } in my code?08:41
ajalkaneyes lard08:41
lardmanstrange, it doesn't work08:42
lardmanah, might be an anchors problem...08:42
lardmanidentical code inline and in the separate file and the latter doesn't work - I get a patterned background08:43
lardmanwhich I guess means it's empty or somesuch08:43
ajalkaneinline inside TabGroup\08:44
lardmanyeah or created inside TabGroup08:44
ajalkaneand is the SummaryWindow also inline in TabGroup?08:44
lardmanIf I create a Page {} inline it works, if it's in a separate file it doesn't08:45
*** antman8969 has joined #harmattan08:45
lardmanseparate file but instantiated inside the TabGroup that is08:45
MohammadAGI didn't know you could create objects inline in QML08:45
lardmancrap, my fault08:46
lardmanmissing anchor08:46
lardmanit would be good to be able to enable borders on the components so I could see them before I populate them and check where they are08:47
ajalkanethe widget gallery app has a tool that enables something like that08:48
ajalkanei havent tested if it works only for mtf apps but also for qml08:49
w00t_it won't work for QML08:51
*** antman8969 has quit IRC08:51
*** piggz has quit IRC08:53
ajalkaneunfortunate. Sometimes I feel like using mtf. Then I look at the example code and go back to qml08:53
lardmanIs it possible to obtain coordinate information from the qml engine for debugging purposes? E.g. if it would spit out a list of objects and their x,y,width,height that would be helful08:54
lardmanhelpful even08:54
MohammadAGweird, now when swiping a windows it becomes transparent08:56
MohammadAGcool bug08:56
w00t_ajalkane: having done a lot of MTF, I have to say that you probably don't want to use it08:56
*** maxw has quit IRC08:56
w00t_you can make nice things, but it's a lot easier to make the same nice things with QML08:56
w00t_and it has a lot of very annoying performance characteristics which involve a lot of working around08:57
MohammadAGMTF in Qt Designer would make it awesome08:57
ajalkanelardman: well, each item has a list of children so you could do it08:57
MohammadAGw00t_, how's the grooveshark client coming along?08:58
lardmanajalkane: I seem to have spent loads of time messing about with my qml code only to find that I was missing an anchor08:58
*** maxw has joined #harmattan08:58
w00t_MohammadAG: temporarily slowed down by real life atm, I have a mountain of paperwork to sort out and a house to redecorate08:59
ajalkanew00t_: I got that feeling too. Its just frustrating qt components miss many components used in bundled apps08:59
rm_you|ah someone is doing a grooveshark client? excellent08:59
w00t_it can (sort of) play tracks, but has a lot of bugs in doing so08:59
rm_you|what are you writing it in? C?08:59
w00t_I also have a half-done lookup client, but I doubt anyone outside of norway will be interested in that... :)09:00
w00t_rm_you|: C++/Qt09:00
rm_you|what backend does it use to play music?09:00
*** smoku has joined #harmattan09:00
ajalkanelardman: I've had some major battles with those anchors too :)09:00
MohammadAGah, k, any git repos around for it?09:01
MohammadAGw00t_, wait, no QML?09:02
w00t_rm_you|: right now, QtMultimedia, but I'll be switching to using gstreamer directly to get working playing from a stream, sometime09:02
w00t_MohammadAG: QML09:02
w00t_I haven't pushed the work I have yet, as it's in a way incomplete state09:02
w00t_I can push it to a branch I guess09:02
w00t_ and for the client09:03
w00t_MohammadAG: pushed to harmattan branch of groovy.. note that it has a bundled copy of libgroove for ease of building09:06
*** xarcass has joined #harmattan09:09
lardmananyone got any tips for working with columns?09:10
lardmanI'm trying to get 3 columns to show up, but I can only see one09:10
lardmanwell the content of one that is09:11
lardmanah, using a row around the columns seems to have done the trick09:13
ajalkaneyou're gonna experience a world of fun using anchors with those. I know I did!09:14
lardmanyeah :(09:15
ajalkaneI won't spoil anything09:15
lardmanI'm at a loss as to how to make the 3 columns I have within my row take up the entire screen width09:15
lardmanfine they will take up enough space to show their content (coloured rectangles for the time being), but that's not ideal at all09:16
lardmanperhaps I should do away with the row and column and just anchor everything together09:17
ajalkaneperhaps. Or you can do parent.width anchoring. Anyway, you're entering the fun-zone then09:18
xarcasslardman: i've found columns pretty useless, unless you're using repeater inside them. anchors are much better. btw, columns are pretty dumb - they can't cope with children that are dynamically change their size or position09:19
lardmanI'm not sure I'm too keen on all the x,y absolute placements going on in the example qml code09:19
lardmanxarcass: ok, that's what I've just realised, glad it's not just me :)09:19
ajalkaneabs placement is an abomination09:19
lardmanbut seems necessary in qml09:20
lardmanalso absolute width and height09:20
xarcasslardman: wrong again09:20
ajalkanei've so far managed to avoid09:20
lardmanI seem to need to set width and height of my components otherwise they default to size 009:21
lardmanthough I guess I could do that dynamically to fit the content09:21
*** DocScrutinizer has quit IRC09:21
xarcasswhy not to use anchors?09:21
*** DocScrutinizer has joined #harmattan09:21
lardmanit didn't seem to work, so I assumed09:22
lardmanbut I may well be wrong09:22
ajalkaneheight I've had to set also... but never position09:22
*** djszapi_1 has joined #harmattan09:24
sandst1lardman: anchors.fill: parent didn't help..?09:25
lardmansandst1: that's fine for a single component, but I need 3 columns in my Page09:25
lardmanso I do anchors.fill: parent for the page, then am anchoring the left and right of each column to one another and to the edges of the parent09:25
lardmanand I still see only one column09:25
MohammadAGty w00t_ :)09:25
w00t_patches welcome09:26
MohammadAGsure, just heading to bed for now09:26
MohammadAGmoo DocScrutinizer09:26
lardmanhi Doc09:27
sandst1lardman: ok. and having a row with anchors.fill:parent and the columns inside the row with anchors.fill:parent didn't work either..?09:27
*** djszapi_ has quit IRC09:27
MohammadAGnight everyone09:27
sandst1lardman: now that i rethink that one, i think thaat won't work :)09:27
DocScrutinizero/ MohammadAG09:27
lardmanI think the last column then fills the area with that one sandst109:28
ajalkanecolumns and rows with anchors equals fun-zone09:28
lardmannight MohammadAG09:28
lardmanI guess I could stick everything in a rectangle instead if that works better?09:28
DocScrutinizerlardman: seems like back when I did such layout in Qt designer ~5+ years ago that's not been any issue to make cols autoresize09:29
lardmanand then along came QML ;)09:30
lardmanreplacing Column with Rectangle of course means I now need to anchor the contents to one another09:31
lardmanspaghetti junction09:31
lardmansurely there's a better way than needing to explicitly anchor everything?09:31
DocScrutinizerOTOH we rarely ever resize on devices/"window managers" like N9(50)/hildon|whatever09:31
artemmlardman:  so you 3 columns to occupy width of the whole parent rectangle?09:32
lardmanartemm: that's what I'd like09:32
artemmand be equal in width?09:32
*** rcg has joined #harmattan09:32
lardmannot bothered about equal width, in fact I'd prefer them to resize according to the content09:32
xarcasslardman: if you have children which geometry changes dynamically, there's no other way09:32
artemmIf their width changed dynamically, how can they fit whole parent width except by accident?09:33
lardmanartemm: the columns default to fitting a 1/3rd of the width unless the content requires that they move09:33
DocScrutinizerchanging width by content is hardly ever working OOTB09:34
artemmthere got to be at least one flexible element that takes "the rest" of width09:34
lardmanas would happen with layouts in Qt/C++ afaik09:34
artemmlardman, what happens in one of the column's content wants to be huge?09:34
artemmjust trying to get the ideal picture of what you want09:34
lardmanin C++, you set a min width for the layout and the other column can grow to fill the screen bar the min space required for the other column09:35
artemmso one of the columns is to be "rubber" and fit the remaining width?09:35
lardmanthat's what I'm used to, just trying to work out how on earth to get something to even display with QML09:35
lardmanall of the columns should do that, and presumably thay is what should happen using anchors too09:36
artemmsorry, still can't get the logic of what you want09:36
lardmanI need 3 columns on the screen09:36
lardmanI can only display one atm09:36
DocScrutinizerwidth = 0.33 unless other col wants >0.33 for huge content, unless own content needs >0.33?09:36
lardmanI can't anchor columns together09:36
*** lamikr has joined #harmattan09:37
ajalkaneyou cant, because row handles their position09:37
artemmajalkane: that doesn't have to be a row09:37
lardmanok, now if I use a row, all my columns sit over on the LH side of the screen, how can I space them out within the row?09:38
ajalkaneo, and shouldnt in this case i think09:38
artemmlardman:  what I am after is the logic of negotiation after some column wants to be bigger or the other one smaller or both09:38
djszapi_1lbt: what changed on obs recently ? kdelibs now seems to build in each repository without issue09:39
sandst1lardman: is the row itself anchored to the center of the screen?09:39
* xarcass wonders what lardman would do if orientation changes09:39
sandst1lardman: as in anchors.centerIn: parent09:39
lardmanartemm: I don't see any logical difficultly there as long as the minimum widths are < screen width?09:40
lardmansandst1: no, to both edges09:40
artemmlardman: you told that u want columns fir whole screen09:40
lardmanartemm: yes09:40
sandst1lardman: try anchors.centerIn instead of anchoring both edges?09:40
artemmif screen is 1000px and columns need 100, 300 and 500, how will they use the screen space09:40
*** elzalem has joined #harmattan09:41
lardmansandst1: ok will give that a go, but I guess then I'll end up with my columns in the centre of the screen, but still not taking up the whole width?09:41
DocScrutinizerlardman: I'd think you could make left of 3 col minwidth=0.25 maxwidth=0.5 and fill the right 2/3...1/2 of screen with a rect taking the rest. Inside this rect you can place the other 2 cols with width=0.5, or you redo the same dance as for leftmost col, to prioritize which col can expand and which element will shrink then09:41
sandst1lardman: yeap..09:42
lardmanartemm: I guess you'd allocate column x offsets as a proportion of the actual content widths and then scale all of the x offsets as any of the contents becomes larger than the allocated region09:42
sandst1lardman: then i guess it's about entering the width: parent.width*x09:42
lardmansandst1: what I'd ideally like is for my 3 columns to be spread across the screen09:43
lardmansandst1: ok, that sounds like it would work, thanks09:43
artemmlardman: Aha! So you want the columns to take amount of screen proportionate to what theur column wants?09:43
sandst1lardman: equally spread?09:43
artemm*to what their content wants09:43
artemmand regardless of absolute width of this content09:43
lardmansandst1: well they all have 50 pixel rectangles in them atm, so they will be equal, but not necessarily once they have something useful in them09:44
lardmansandst1: The problem I see with setting the absolute position as some fraction of parent.width is that then you may end up overlapping things presumably?09:44
lardmanI guess I'll just anchor everything together09:45
DocScrutinizerlardman: doesn't this get handled by setting relative fractions of width as a pos?09:45
sandst1lardman: i'm not 100% sure about overlapping, it could be also that the children in the row get only as much of the space as there's left09:46
lardmanDocScrutinizer: that would position them as I'd like, but then if one of the columns is > parent.width/#cols it will encroach on the others I expect09:46
DocScrutinizersounds like QML is really terribly limited in what functions it provides to relieve pain of layout from developer09:46
xarcasslardman: if you'll decide to use these fractions, don't forget to Math.round() their results. otherwise, there will be unpleasant artifacts09:47
lardmanxarcass: ok thanks09:47
lardmanDocScrutinizer: Well it seems very flexible, I was just hoping for some shortcut layout type elements to save me needing to do lots of anchoring while I'm testing code09:47
xarcasslardman: use copy/paste :)09:48
lardmanxarcass: need to change the id I'm anchoring against though of course09:48
ajalkaneset widths, but position with anchors09:49
* xarcass remembers that video lecture about QML, where charismatic bald man said that it's ok to use copy/paste in QML coding09:50
* lardman goes to find caffeine before trying to write out the anchors between his 15-odd ui elements on this page09:50
artemmwidth + positioning with anchors is easy. Doing that right now trying to get the column resizing logic. Everything works, but I still don't get the logic you want :)09:50
*** veskuh has joined #harmattan09:50
artemmok, I give up probably09:51
artemmcan't understand without a picture09:51
DocScrutinizerlardman: sorry, seems I didn't manage to get what's your design goal. Anyway there's either just one "rubberband" element in your layout, and that takes all of the remaining space, or you got several rubberband elements and aiui those will all be same width then09:51
DocScrutinizerand I dunno how to make width of main window increase when sigma_width of widgets in window is too large for the current window width09:52
lardmanNo worries chaps, I'll do some anchoring now I have a cup of coffee to hand :)10:01
lardmanand thanks for all the suggestions and help :)10:01
DocScrutinizerlardman: basically my approach (NB while I got NFC about QML) would be to start up Qt Designer on my PC and click together a layout that meets my needs. Then check which fields of width_relative and width_abs and also posX_relative and _abs have which values. If QML doesn't support these values/semantics OOTB you need to code the underlying algorithms "by hand" (e.g if there's no posX_relative then use posX_abs=window.width()*$X_10:01
*** harbaum has joined #harmattan10:02
lardmanDocScrutinizer: The thing is I'm not quite sure how I'll handle the column widths and positions anyway, I just wanted to space them out so I could test and add in components and ensure that I can see them/check whether I need to set widths/heights10:03
lardmanI'll probably end up with one column being partially offscreen then being flick-on-able10:03
lardmanotehrwise the content will be too wide10:03
DocScrutinizeryou can easily prototype and test different layout ideas in Qt Designer10:04
DocScrutinizerif you're happy with the result, transfer it to QML10:05
DocScrutinizer(of course unless there's maybe a similar too for genuine QML :-D )10:05
lardmanDocScrutinizer: yeah, I've got most of the UI stuff sketched on paper, which is as good as10:05
xarcasslardman: how do you think to cope with orientation changes then?10:05
DocScrutinizers/too /tool /10:05
lardmanxarcass: I'll change anchors, but had hoped to only change anchors on my columns10:05
infobotDocScrutinizer meant: (of course unless there's maybe a similar tool for genuine QML :-D )10:05
*** lcuk has joined #harmattan10:07
*** lcuk has quit IRC10:07
*** lcuk has joined #harmattan10:07
lardmanIt's my fault really, I'd not realised that absolutely everything requires explicit anchors10:07
xarcasslardman: a lot of work lies ahead of you. especially if all this is placed inside a flickable. have you considered using tabs?10:08
lardmanxarcass: I am using tabs already10:08
lardmanxarcass: but yeah, I realise there's some work to do, but it should be pretty easy to get the anchors right, just will take some time10:09
xarcasslardman: then there's a lot of content in your app..10:09
lardmanI'll have a column with scanner barcode image and under that the type of barcode and scan date, centre column has some buttons to copy&paste, save to file, web search and underneath is the plugin supplied "widget" and right and column is the plugin list that alters what is displayed in the centre column10:10
lardmanthat would be landscape, portrait some of the columns will move to become rows, etc10:11
lardmanhmm, not the best spelling in the world there, hope it made some sense :)10:14
xarcasslardman: that's not so much as i've thought. can't see any tabs though10:14
lardmanI've got tabs for the history list and settings pages10:14
xarcasslardman: my point was that this design isn't too complicated as i suspected. actually, it's piece of cake. i'm sure you'll manage :) moreover, afterwards you'll remember present difficulties with laughter10:17
lardmanxarcass: indeed the design is very simple, which is why it's so frustrating not being able to actually get much to display on screen :)10:18
lardmandoes one need to set the top and bottom anchors of touching elements or is setting just one enough?10:19
xarcassxarcass: i clearly remember how pathetic i was when i began QML coding. it'll pass soon enough. keep trying10:19
lardmanyeah I know, just fruistrating as this should be the easy bit :)10:19
xarcasslardman: it's either top anchor and height, or both top and bottom anchors10:20
lardmanbut for both elements which are touching?10:20
lardmanas the top anchor of the bottom element is just repeating the bottom anchor of the top element10:21
lardmanrepeating the link anyway10:21
xarcassusually, when i fill a column, i set top anchor and height for all elements which height can be calculated10:21
lardmanok cool10:21
*** Mek has quit IRC10:26
*** khertan has joined #harmattan10:27
* lardman cries10:27
*** Mek has joined #harmattan10:28
lardmanand curses his Rectangles which don't want to line up, or even appear other than the first one10:28
marioblardman: You just have to make sure you have a width and height for Rectangles. Either by anchoring or by settings width/height10:29
lardmanah, missing anchors on the contents10:29
lardmanyeah I set width/height but forgot to anchor these ones which had been commented out while I was testing a single column10:29
marioblardman: :)10:29
lardmancurious that the rectangles don't sit at their parent's origin though10:30
lardmanthough perhaps my anchoring of the fake-column rectangles is also broken10:30
marioblardman: Inspect the output carefully and look for anchoring loops and other errors10:32
lardmandefinitely no errors in the output10:33
*** djszapi_1 has quit IRC10:33
lardmanit seems my fake-column rectangles aren't anchoring to one another, they all just sit at the origin and draw ontop of one another10:33
xarcasslardman: you can easily check this out by putting console.log("x" + x) in some hook, like Component.onCompleted10:36
lardmanxarcass: I've got coloured squares being drawn so can see it's not working, but thanks10:36
lardmanI'll just pastebin it and if anyone can tell me why the Items (which were Rectangles, but as they are just layout elements that was overkill) don't anchor that would be great10:37
* xarcass also likes colored rectangles. especially when debugging geometry issues10:37
lardmanall the columns/Items seem to draw, but all of the columns are drawn stuck to the LH side of the screen10:41
lardmanso it seems the anchoring of the Items isn't working for some reason10:41
xarcassanchors { left: parent.left; top:; } - that's in the root Item. now try to guess which would be the height of that Item?10:41
lardmanpresumably it will be as high as the content10:42
lardmanwhich is what happens on the screen anyway10:42
xarcasswrong. correct answer is zero10:42
lardmanodd, because that column is drawn10:42
xarcasscan be easily checked by adding "clip: true"10:42
xarcassdon't forget that clipping is disabled by default10:43
mariobxarcass: Me too remember the old days with height/width problems :)10:43
lardmanok, so I'll add  bottom: parent.bottom to all of the Items10:43
lardmanbut still they are all drawn over the first column10:44
xarcassit's not related to geometry problem, but nonetheless:                     fillMode: Image.PreserveAspectFit - this has meaning probably only if you set Image size directly10:44
lardmanxarcass: I've not got round to testing that yet, need to get something to display first ;)10:45
lardmandoes one need to set the size of the image element? It doesn't take it from the source image file?10:45
artemmit should10:45
lardmanThe image will only be a thumbnail anyway, so I'll need to set either the width or height10:46
xarcassclear. your columns overlap because they not only have zero height, but zero width also10:46
*** djszapi has joined #harmattan10:46
lardmanhmm, so they don't take on the width of their child elements?10:47
lardmanI'm too busy thinking of these things as layouts rather than as elements in their own right10:48
lardmansilly me10:48
lardmanthanks, that should solve quite a few problems then10:48
xarcasslardman: you have two import QtQuick lines - with different versions at that. beware: if you are using two different versions of QtQuick, even if they are in different QML files, your app's memory footpring may increase by 100+M. i've seen this myself10:51
lardmanI was copying one of the example files that did that10:51
djszapihiemanshu: did the import work on N950 ?10:51
lardmanbut I'll remove version 1.010:51
lardmanthanks xarcass10:52
*** andre__ has joined #harmattan10:55
*** andre__ has joined #harmattan10:55
*** frinring has joined #harmattan10:59
*** wazd has joined #harmattan11:08
*** mzanetti_borg is now known as mzanetti_work11:09
*** leinir has joined #harmattan11:22
rcgwhat's the best way to install additional -dev packages into the QtSDK Madde sysroots?11:36
rcgah.. that should do the trick11:40
*** achipa has joined #harmattan12:07
*** M4rtinK has joined #harmattan12:10
lardmanHas anyone noticed strange issues with text.width?12:13
lardmanI've got a qml file to create a button, and it specifies: height: text.height + 10; width: text.width + 2012:14
lardmanbut the buttons seem to be far larger than that12:14
*** djszapi has left #harmattan12:14
lardmanah, it's not using my Button implementation, that could be the reason why12:15
*** smoku has joined #harmattan12:15
*** lamikr has quit IRC12:18
*** Arkenoi has quit IRC12:23
*** Arkenoi has joined #harmattan12:24
*** sudanix has joined #harmattan12:38
*** lamikr has joined #harmattan12:38
*** lamikr has quit IRC12:49
*** lamikr has joined #harmattan12:49
*** lamikr has quit IRC12:50
artemmHmm, why am I seeing qDebug() output when I choose Harmattan Release target in QtCreator?12:55
artemma bug? Or am I misunderstanding how qDebug works?12:55
artemmsame about QML/JavaScript's console.debug() actually12:55
artemmMeeGo 1.2 Harmattan API, not Platform API for Harmattan if it matters12:56
artemmqmake build steps (from Projects tab) don't include CONFIG+=debug12:58
rcg"... This function does nothing if QT_NO_DEBUG_OUTPUT was defined during compilation. ..."13:00
rcgfrom the qDebug api docs13:00
leinirartemm: Even while your own code might be release compiled, Qt can still be compiled with debug :)13:00
rcgseems like you have to explicitly disable debug output13:00
leinirAh right :)13:00
artemmdoes it mean that N950 includes debug version of Qt only?13:01
artemmI am seeing these debug prints from HW13:01
rcgafaiu enabling/disabling debug in the build via e.g. CONFIG+=debug only triggeres whether your project is build with debug symbols enabled13:01
artemmhey, I am seeing my own debug prints, so doesn\t it mean that my code is debug built?13:01
rcgartemm: afaik the -dbg packages are not installed by default13:01
rcgthere are *-dbg packages available though13:02
artemmwell, when I choose Release build, CONFIG+=debug is not included13:03
rcgwell.. i dunno if _not_ setting CONFIG+=debug implicitly means QT_NO_DEBUG_OUTPUT is passed during compilation13:03
artemmthat's what Project tab shows13:03
artemmwhere do you put QT_NO_DEBUG_OUTPUT?13:03
artemmto qmkae arguments?13:03
ajalkaneExcuse my manners13:04
artemmwhat I am actually worried about is that I can't see whether my build is really release or debug one :) Seeing debug prints is one of the signs only13:05
artemmI don't need debug prints and debug symbols in the final build13:05
*** khertan has quit IRC13:06
rcgartemm: hmm yeah, ic13:07
artemmaha, QT_NO_DEBUG_OUTPUT goes to DEFINES in the .pro file13:07
artemmI have small hope that it's not my code is so slow, but it is just compiled for DEBUG, while I thought it's RELEASE :)13:08
ajalkaneIf theres lots and lots of debug logging, it'll slow it down13:10
artemmActually any app launch prints some 10-20 lines about "MeegoGraphics", found drivers, etc. - that is obviously coming from the system13:10
artemmAdded DEFINES += QT_NO_DEBUG_OUTPUT to pro file, ran qmake, rebuilt, still seeing my qDebug() prints13:11
rcghmm artemm you are right.. just checked here.. even in release "-g" is passed as g++ command line option13:15
artemmsorry, I was wrong. Proper rebuild removed qDebug prints13:15
artemmconsole.debug are still printed though13:15
artemmwhat is -g?13:15
artemmalso that QT_NO_DEBUG_OUTPUT removed my own debug prints, but not the system ones. So I assume I am still linking to debug libs somehow13:16
*** meegomy_ has quit IRC13:16
rcg".. -g Produce debugging information in the operating system's native format .."13:18
artemmreading this..13:18
artemmthat is still about my own code, isn't it?13:18
*** veskuh has left #harmattan13:18
rcgyep.. so essentially afaik "-g" means g++ is adding debug stuff to a build13:19
artemmhmm, I don't see any -g in the qmake line13:19
artemmHere's "effective qmake call":13:19
artemmqmake /Users/artem/WorkArea/Dropbox/discountcalc/trunk/discountcalc-harmattanapp/ -r -spec linux-g++-maemo -unix13:19
*** lamikr has joined #harmattan13:19
rcgnot in the qmake call13:20
artemmalso my app is 95% QML + JavaScript, so I should be more worried about linking to release libraries, shouldn't I?13:20
rcgin the g++ calls13:20
rcgwell dunno how this is handled in the QML part13:20
artemmas qml is interpretable anyway and release optimization could maybe kill qDebug prints and that's it13:20
artemm*searching Compile output for g++ call13:21
artemmindeed -g is passed there13:21
artemmhow can I control it?13:22
rcgheh.. dunno..13:22
rcgalways thought switchin between "release" and "debug" would take care of it ;)13:23
artemmme too :)13:23
rcgactually never looked that precisely at what its really doing13:23
artemmwell, there's hell a lot of problems with default project properties set by QtCreator harmattan wizard13:23
artemmI guess just nobody had a serious look at it13:23
artemm*went to read about QMAKE_CXXFLAGS13:24
rcgwell.. just had a look at my build in scratchbox.. this also builds with "-g" set13:25
tommaCONFIG-=debug CONFIG+=release should do the trick13:25
artemmthere seems to be QMAKE_CXXFLAGS_RELEASE flag13:25
artemmthanks, tomma, will try it now13:26
artemmnope, -g is still passed to g++13:27
rcgtomma: just checked.. it doesnt13:27
rcgstrange thing13:27
rcg-g  Produce debugging information in the operating system's native format13:28
rcgjust to have the correct source ;)13:28
artemmsomebody had similar problems -
artemmnot that I understood how to use the answer though13:28
rcgwell.. i really need to hit the road...13:29
artemmit's a pity13:29
artemmgood luck with your way anyway13:30
artemmmaybe somebody else can help13:30
tommadoes it give -02 to g++ ?13:30
rcgthanks.. i hope you get your issue sorted out as well :)13:30
tommathen it is building release with debuginfo... no idea why13:30
artemmtomma, yes it does13:31
tommawell it can be removed with QMAKE_CXXFLAGS -= -g13:31
tommai hope13:31
xarcassi suppose, in release version it just strips binary before adding resources. in my case, binary is 300k in release and 6M in debug builds13:31
artemmJust added QMAKE_CXXFLAGS -= -g to pro file, ran qmake, started build, still seeing -g13:32
xarcassand yes, it has -g flag in release. and -O2 as well13:32
artemmwrong var13:32
artemmQMAKE_CXXFLAGS_RELEASE -= -g helped13:33
artemmthere's no -g in g++ call anymore13:33
artemmshould I also remove -O2?13:33
xarcasswhat would be the point?13:33
artemmI am not sure which exactly flags to care about for producing fastest non-debug build13:34
artemmwell, I don't really know qmake13:34
tomma-02 is for optimizing13:34
xarcassremoving -g won't help here13:34
artemmtrying to optimize the final build13:34
artemmfunny, package creation seems to be requesting -g from the system13:35
artemmdpkg-buildpackage: set CFLAGS to default value: -g -O213:35
artemmdpkg-buildpackage: set CPPFLAGS to default value:13:35
artemmdpkg-buildpackage: set LDFLAGS to default value:13:35
artemmdpkg-buildpackage: set FFLAGS to default value: -g -O213:35
artemmdpkg-buildpackage: set CXXFLAGS to default value: -g -O213:35
artemmdoes it mean it links to the debug libraries>13:35
tommait links to same libraries but includes debug symbols in binary13:35
xarcass-g adds debug info. which will be stripped in release anyway13:36
tommabut doesn't it strip binary in packaking?13:36
artemmwell yeah, packaging packages already compiled code13:36
artemmwhat if I don't want the debug symbols?13:36
artemmthe real question is how to make my final build the fastest possible. Questions about debug symbols are supposed to help me get there13:37
xarcasstry debug build, look at the size of the package and feel the difference. there's no debug symbols in release binary, i'm almost sure13:37
*** baraujo has joined #harmattan13:38
xarcasswell, there are many ways to make things faster here: less JS, more C++. and QML booster, maybe13:38
artemmtrying debug build now13:38
artemmqml booster is used already13:38
artemmwell, current speed isn't that bad13:39
artemmbut if I want to improve it by playing with config options, why not :)13:39
artemm*if I can13:39
xarcassone more thing: guys from nokia office, to whom we are sending our packages, say that it works much faster on the N9 than on the N95013:40
artemmwhat could be a reason?13:40
artemmjust fresher build?13:40
artemmor image compiled with release libs?13:40
xarcassvery likely13:40
artemmthat would be good13:41
villagerartemm: the debug symbols added by -g has zero, *zero*, effect on performance as they're not used unless you're debugging, but it is good practice to always include it in the build... though the packaging process removes the symbols anyway for size13:41
artemmright now when starting app on N950 I see quite many system debug messages13:41
villagerartemm: but using -O2 (or even -O3) is important for good performance...13:42
xarcassvillager: can't agree with you: adding these symbols significantly increases deployment time, which isn't convenient13:42
villagerxarcass: I doubt that matters after packaging, since they're stripped away then13:43
villagerxarcass: and before, they're pretty useful if you need to get a backtrace or whatever13:43
villagerxarcass: anyway I'm talking runtime performance, of course13:43
xarcassmy apps do not produce backtraces :)13:43
artemmmade release and debug build to compare sizes13:46
artemmdifference in the final .deb package is 6 bytes..13:46
* xarcass just tried debug build and is very surprised with all these 500 warings13:46
artemmnot a whole lot of binary optimization :)13:46
lardmanhmm, my symbols weren't being removed13:46
mikhas_xarcass, artemm: take note that QML models tend to be on the inefficient side of things13:47
xarcassthen I was wrong about size13:47
mikhas_always querying a full blown QObject for example, for each model.get13:47
xarcassbut I can remember that i've seen this. can't remember when, though13:47
artemmMikhas_, true, but 6 bytes of removed debug symbols?13:47
mikhas_(or creating one, for model.set)13:47
mikhas_so if you want to speed up your views, consider optimizing the models13:48
xarcassyes, i've remembered, that was before it was optimized for the booster!13:48
artemmthere's very small amount of models in my app13:48
artemmquite a lot of property binding during start though13:48
artemmthat is something to optimize too13:48
artemmbut compiler flags are easier to try :)13:48
*** rcg has quit IRC13:48
artemmit's a pity qml analyzer isn't released yet13:50
artemmtried it about a month ago on some own-built version of QtCreator, didn't work at all13:50
artemmto close the topic about linking optimization: should I be worried seeing this kind of messages during startup:13:51
artemmMeegoGraphics: found EGL_NOK_image_shared13:51
artemmIsn't it something system should NOT tell in the release build?13:51
w00t_your application may be built in release mode, the library telling you that is not13:51
artemmsorry, don't understand :/13:52
w00t_your application uses libraries, such as Qt13:52
w00t_when you build your application, you don't build Qt13:52
artemmyou mean I am using the debug version of library?13:52
w00t_so, building your application in release mode doesn't affect how Qt (for instance) was built13:52
artemmok, then is it possible to link against the release version of Qt?13:52
w00t_you are, it just has debug messages enabled13:53
artemmI used to do similar stuff at Symbian times, but Qt world is still new to me :)13:53
w00t_that is, it has been built without debug symbols, with optimisations, but with qDebug() enabled13:53
artemmthen the final devices (in consumer hands) will probably have all this stuff optimized firther and running faster, right?13:54
mikhas_the final devices will poop double rainbows13:54
*** macmaN has quit IRC13:54
artemmI don't care about rainbows unless they are pooped with 60fps at least ;013:55
w00t_the particular bug that causes this to be neccesary is complicated, so I doubt it'll be fixed in time, and besides that, the few debug messages that come from Qt itself won't affect performance13:55
lcukmikhas_, photos or it didn't happen!13:55
lcukI thought n9 final devices were tri coloured :P13:55
artemmi guess it means that performance I see on N950 will be similar to final N9 (maybe jsut a little faster)13:56
artemmthanks for your patience13:56
mikhas_well, core apps are still being optimized13:57
w00t_well, the software that will be released will be newer (and different, ideally a lot better) from what you have now13:58
mikhas_for 3rd party apps this means: we leave more CPU juice and memory for you13:58
*** lcuk has quit IRC13:58
artemm*now went to search why javascript's console.debug() is printing when qDebug() doesn;t13:58
spenapartemm, maybe you have to run your application with -output-level debug ?13:59
*** lamikr has quit IRC14:00
mikhas_wasn't that only for LMT apps?14:00
artemmwell, I want to disable console.debug() prints in release build14:00
w00t_mikhas_: right14:00
*** macmaN has joined #harmattan14:00
artemmBTW, does anybody know roughly how long a printing a line takes? Around 20-50ms?14:00
artemmthinking abt how strict I need to be removing the debug prints14:01
w00t_much less than that14:01
w00t_much much less14:01
w00t_it also depends on where they're going14:01
artemmok, thanks14:01
artemmthen it's not so critical, I have just a few14:01
artemmmaybe 10-15 lines during startup14:02
w00t_console.debug can't be disabled, btw14:02
w00t_(I just checked)14:02
artemmfunny, why does it exist separately from console.log then? :)14:02
w00t_it's tied to qDebug14:02
artemmoh well, it's probably javascript implementation14:02
w00t_so it *can* be disabled, but not without rebuilding Qt14:02
artemmw00t_: on C++ side of my code qDebug doesn't print anything (disabled by QT_NO_DEBUG_OUTPUT or QMAKE_CXXFLAGS_RELEASE -= -g)14:03
artemmand Javascript side still is able to print via console.debug()14:04
w00t_remember what I said about libraries14:04
w00t_this use of qDebug is inside Qt itself, which isn't built with QT_NO_DEBUG_OUTPUT14:04
w00t_'this use' being console.debug()14:04
artemmthen why qDebug's on C++ side of my code can be disabled?14:04
w00t_qDebug is entirely inline14:06
w00t_i.e. contained within headers14:06
artemmok, got it14:06
w00t_when you define QT_NO_DEBUG_OUTPUT, Qt turns it off for where you've included it, i.e. your code14:07
*** lamikr has joined #harmattan14:08
*** rcg has joined #harmattan14:15
*** lcuk has joined #harmattan14:20
*** lcuk has quit IRC14:20
*** lcuk has joined #harmattan14:20
*** lamikr has joined #harmattan14:27
*** willer_ has joined #harmattan14:31
*** NIN101 has joined #harmattan14:49
*** berndhs has joined #harmattan14:50
artemmIs anybody able to debug on N950?14:53
artemmWhen I "Start Debugging" I get The gdbserver process closed unexpectedly.14:54
artemmsoon after "Can't bind address: Address already in use."14:54
*** CepiPerez has joined #harmattan14:57
xarcasswlan or usb?15:01
artemmbut maybe I was too quick15:05
artemmjust figured qml debugging library wasn't built15:05
xarcasstry to enable 'background connections'15:05
artemmmaybe that's the reason15:05
mikhas_use the USB connection15:07
artemmnope debugging lib didn't help15:07
artemmwhat is backgroundconections?15:08
artemm*went for usb cable15:09
*** Arkenoi has quit IRC15:16
artemmusb didn't help15:17
artemmsame result15:17
xarcassi have just tried debug - works perfectly with breakpoints and all that stuff. through USB.15:22
xarcasstry to enable background connections and ensure that power saving is disabled15:22
*** NIN101 has quit IRC15:23
artemmwhat are these background connections?15:25
artemmin device?15:25
artemmin qt creator?15:25
xarcassyes, in settings->internet connection15:26
xarcasshowever, i doubt that it'll help with USB15:27
artemmhad to disable flight mode to see background connections setting (discounnted from wlan manually)15:27
artemmsame result :/15:28
artemmok, will try over wlan then :)15:28
xarcassit was in the flight mode?15:28
artemmto connect via usb you need to disable wlan15:28
artemmconnection to qt creator was still happening15:28
xarcasssomething is wrong with your setup. i can use both simultaneously. do you  have latest firmware?15:29
artemmheh, I think I've seen debugging working fine about a week ago15:30
artemmnot sure as I didn't really use it, just was playing with the device15:30
artemmlatest public n95- fw15:30
artemmn950 fw15:30
xarcasshave you flashed it yourself? with that one-click-something?15:31
artemmmaybe I just restart everything15:31
artemmwith one-click-something15:31
artemmabout a week or two ago15:31
artemmCould use of booster influence the process somehow15:32
xarcasswell, reboot also helps sometimes. apart from that there's nothing more that I can offer, unfortunately15:32
*** trx has quit IRC15:32
artemmI think it's only wholeapp level thing i changed lately15:32
xarcassmy app is also boosted, but debugger works nevertheless15:33
artemmdoes QML analyzer works for you?15:33
artemmthat is the actual reason why I want debugging15:33
artemmso if it doesn't work yet, no need to struggle with getting debugging started15:34
artemmsurprise-surprise restart helped :)15:36
artemmand starting of debugging takes a loooot of time15:37
artemmso many slow messages about loading libraries15:37
*** trx has joined #harmattan15:37
*** lcuk has quit IRC15:40
artemmdebugger is "setting breakpoints" for about 10 min already. Is it normal?15:43
*** Evgeniy has quit IRC15:44
xarcassthat's normal15:45
artemmgosh :(15:45
artemmok, now I remember why I didn't use debugger earlier :)15:46
artemmnever was patient enough15:46
*** lcuk has joined #harmattan15:47
*** lcuk has quit IRC15:47
*** lcuk has joined #harmattan15:47
*** javispedro has joined #harmattan16:01
*** hardaker has joined #harmattan16:04
artemmIs it possible to develop for Harmattan using QtCreator 2.3 (in release candidate for now)16:08
artemmHarmattan plugin works nice for 2.2 on my computer, but I am not sure how to add it to 2.3 if it's possible at all16:09
*** crevetor has joined #harmattan16:10
*** crevetor has quit IRC16:13
*** crevetor has joined #harmattan16:16
radiofreejavispedro: your qsparql bug has been fixed, it's just not in the beta sdk16:16
artemmoh, 2.3 can open 2.2. projects and all the settings. nice16:16
javispedroradiofree: nice to know!16:16
radiofreeversion 0.0.30 has the fix, also, it's import QtSparql 0.1, not 1.016:16
radiofreeQtSparql 1.0 may work, but that's a bug in qt16:16
javispedroradiofree: so, can it be used, or it has any other showstopper bugs?16:16
radiofree0.1 will work in scratchbox and on device for sure16:16
radiofreeyes it can be used, along with the endpoint driver16:17
javispedromany thanks!16:17
radiofreewhich doesn't work on the device in 0.0.27 (the version in the beta sdk)16:17
radiofreein a future release there will be a whole new set of qml bindings (will maintain this SparqlResultList for compatibility though)16:18
radiofreeso you'll be able to issue SPARQL queries directly in QML, and use the results there :)16:18
radiofreevar results = for example16:18
radiofreebut in terms of list models, SparqlResultList is fine16:19
javispedrothat's mostly what I want16:19
javispedroI'm doing an app that is mostly a list of results from a tracker query,16:19
radiofreethe problem with SparqlResultList is that it doesn't have a get() function16:19
javispedroalong with maybe a thumbnail extractor (I was considering putting it in tracker itself but it's not that used)16:19
radiofreebut there should be other ways around that16:20
artemmCould not connect to the in-process QML debugger:16:21
artemmThe application is not set up for QML/JS debugging.16:21
artemmWhat could that mean?16:21
artemmit is a debug build, qml debugging library is compiled, "link to qml debugging library" is set, debugging type is set to "qml only". What else does it need?16:22
artemmIt's QtCreator 2.3 RC if it matters16:22
*** qgil has joined #harmattan16:23
javispedroradiofree: feel free to close it  then (or should I do that?)16:28
xarcassartemm: i suppose, that would require presence of that qml debugging library on the device. in short: won't work16:31
artemmwhy wouldn't it exist on device?16:31
xarcassbecause our fw is fairly old, that's why :)16:32
artemmI though this "qml debugging library" is statically linked and deployed together with my app. Otherwise why would I need to build it from qt creator16:32
artemmoh well, I believe you know better16:33
xarcassthis may be the case. i don't know exactly16:33
artemmthen no qml profiling until next fw for me16:33
radiofreejavispedro: keep it open, I'll comment on it16:33
xarcassartemm: have you tried analyzer? it's grayed out for me16:34
radiofreejavispedro: if you fancy building it yourself has all the fixes16:35
javispedroradiofree: thanks for that too, was going to do it :)16:35
RST38hg'day javispedro16:35
javispedromorning RST38h16:35
*** faenil_ has joined #harmattan16:35
faenil_challenge completed!16:36
faenil_from 0.2fps to 20fps :D16:36
faenil_hoped to get to 60, but anyway XD16:36
javispedroa 100x increase? what were you doing, sending the rendered frames via a web service or what? :D16:36
faenil_well I have16:37
faenil_10k floor tiles16:37
faenil_and 2k boxes16:37
faenil_in this testcase16:37
faenil_at 0.2 I was using a drawcall for each floor tile and for each box :)16:38
elpurifaenil_: next up occlusion culling =)16:38
faenil_Now I'm using a VBO for the floor, and another VBO for boxes :D16:38
faenil_elpuri: don't think this is the case, because I want the user to be able to look at the whole level, so it must be smooth without culling too :)16:38
artemmxarcass, I can start analyzer from Qt Creator 2.3 RC. Well, I can try starting it as debugging session start can't be completed16:39
elpuriyou've got 10k tiles visible on the screen at all times?16:39
elpuriwow.. ;)16:39
faenil_nope, this is an extreme test case :)16:40
faenil_but I want the user to be able to look at the whole level while having a nice framerate16:40
faenil_so I have to make it smooth without culling too...16:41
xarcassartemm: so, it doesn't work yet. that's a shame. i have to wait more, then16:41
artemmI can't make it work, but it's not the analyzer16:41
artemmI can't start ondevice debugging16:41
artemmfrom Creator 2.316:41
artemmif you can do it, you may be able to use analyzer as well16:41
faenil_anyway, I have named it, it's Qtoban! :D16:41
xarcassartemm: i can't afford such experiments as installing RC versions16:43
artemmcreator isn't whole Qt16:43
artemmit can easily coexist with another instance16:43
*** lamikr has quit IRC16:43
xarcassartemm: i've tried once or twice, it screwed up all my setup. so i don't even think about this now16:44
artemmwell, not even trying is certainly safer :)16:44
xarcassartemm: tight deadlines, unfortunately16:46
artemmand work hard for those deadlines! I want better firmware soon! :)16:46
elpurii wish people would stop naming their apps q<something>16:47
xarcassartemm: that's not my responsibility. i'm just app coder16:47
lardmanelpuri: iSomething instead? ;)16:47
elpurinot that either ;)16:47
elpuriunless you're apple16:48
lardmanThere was the same issue with mSomething for Maemo apps16:48
artemmelpuri, isn't it like a developer-type signature? Kind'of "normal users, please ignore my app, it has a developer-ui and normal people can't use it" :)16:48
kimjug<something>, k<something>, m<something> for gnome, kde, maemo/meego..16:48
lardmanIt does mean you can use a noun and not bother to think to hard16:49
lardmanat least that's my excuse16:49
faenil_elpuri: I hate that too...but looking for a nice name is so hard!16:49
faenil_elpuri: my problem is that my game is a clone, so I can't give a whole new name afaik...or can I?16:50
elpuriartemm: exactly. i think the goal here is to try to give meego/harmattan a credible app portfolio, hence the names should be "normal" imo16:51
elpurifaenil_: why not16:52
elpuricall it tilepalooza16:52
faenil_elpuri: don't know, licence something?16:52
artemmelpuri, only if you want to create apps for normal people :)16:52
faenil_not bad XD16:52
artemmnot everybody cares16:52
artemmany sane product manager would instantly kill any qApp name16:52
artemmas it says nothing to end user16:53
faenil_ok I have time to think about another name16:53
faenil_I just needed one to create my local repo16:53
faenil_to track the changes :)16:53
elpuriartemm: exactly16:53
faenil_I Should find something like StockThemUp16:54
faenil_don't know XD16:54
faenil_WarehouseKeeper, simple16:54
faenil_no, people would misunderstand it16:54
xarcassfaenil_: what's the game like?16:55
elpurisokoban clone?16:56
faenil_it's sokoban clone16:56
faenil_in 3D16:56
xarcassfaenil_: Angry Boxes16:56
faenil_ahahah yeah xD16:56
faenil_that's something people would download :D16:56
faenil_and then be disappointed about xD16:57
xarcassfaenil_: then don't disappoint them16:57
faenil_xarcass: I'll try to, but I'm just learning OpenGL :) and doing this game for a university exam too :D I have 10 days left to finish it XD with a level editor16:57
faenil_omg don't let me think about it XD16:58
faenil_anyway thanks for the names :D16:59
* xarcass has just received confirmation from nokia that all application bugs are fixed at last. it's time for some vodka now17:00
faenil_yey :)17:01
faenil_is there any news about a fw update for us mortals?17:02
*** Kypeli has quit IRC17:02
xarcassfaenil_: it's just an app, that would be preinstalled on the devices in our region. nothing special. it's definitely not a fw, not even a part of it17:03
faenil_xarcass: I did not mean to refer to your msg :D17:03
faenil_I was asking about our N950 fw update :D17:03
faenil_the actual fw is bugged enough :D17:03
*** Kypeli has joined #harmattan17:06
faenil_The internal deadline for the release version of the FW should be 29th..17:09
faenil_so I hoped we were getting some love :D17:09
*** NIN101 has joined #harmattan17:14
*** maxw has quit IRC17:15
faenil_elpuri: is there anything I can do to speed up thing like disabling the compositing?17:18
*** xarcass has quit IRC17:18
javispedroif you are fullscreen it is already disabled17:19
faenil_oh ... ok then...17:20
faenil_swiping away is very slow :D17:21
mariobfaenil_: disable "Developer Mode" increases performance, according to MeeGoExperts,17:21
faenil_something ...17:22
faenil_gained 0.2 fps17:22
javispedroeven more doubtful17:22
faenil_I just tried XD17:22
faenil_and gained 0.2fps17:22
thpwazd: ping17:23
javispedroyour testing method is flawed17:23
faenil_but 0.2 is almost nothing...17:23
MohammadAGdeveloper mode is a metapackage17:23
javispedrofaenil_: it should have shown _no difference_17:23
MohammadAGit can't affect performance17:23
faenil_ok, trust you :D17:24
faenil_I'll try again17:24
MohammadAGit could be that tracker or something like that was running before he disabled it17:24
elpurido you receive any signal when the swiping starts?17:24
javispedrotechnically yes17:25
elpuriyou could take screenshot and render just that paused state while user is swiping17:25
MohammadAGjavispedro, inside QApplication?17:25
javispedroMohammadAG: an atom is set17:25
javispedroa property17:25
faenil_but that's the last problem actually :) I have to pause the game too, because otherwise the phone is very slow17:25
javispedrotherefore you should get a Propertynotify17:25
MohammadAGah, so changeEvent() shoild do it17:25
javispedroI'm, my atom cheat sheet is missing this one17:26
MohammadAGwhen an app is thrown in the background there's a property set17:26
javispedro ;)17:26
javispedrothe arrows in the first column indicate who sets the property and who reads it17:28
javispedroMTF = meegotouch, COMP = mcompositor, swipe = propietary swipe code17:28
MohammadAGgood, now make that in qml so it can have arrows with transitions!117:28
*** cpscotti has joined #harmattan17:29
mikhas_you know, you could have just asked17:29
mikhas_but I guess it was more fun this way17:30
elpurijavispedro: you've got a great letter e17:30
MohammadAGblinking 2G icon, wtf does it mean?17:30
javispedromikhas_: yes :)17:30
MohammadAGthere goes my operator logo17:30
javispedrothe atom you're looking for the swipe starting is _MEEGOTOUCH_MINIMIZE_ANIMATION17:33
* javispedro adds to cheat sheet17:33
lardmananyone know if it's possible to set the anchors of one element to be the same as another?17:34
lardmanI've tried anchors: blah.anchors17:35
lardmanbut apparently that's not allowed17:35
MohammadAGlardman, set them manually to what the anchor of the first one?17:37
lardmanMohammadAG: yeah I need to display a placeholder if I don't have an image, so want it to sit exactly where the image is17:38
lardmanand to be anchored identically17:38
lardmanI've done it manually now, written out each one of top, bottom, left, right17:38
MohammadAGlardman, anchors.fill: firstone17:39
lardmanah, cool17:39
MohammadAGplaceholder, achors.fill: notplaceholder17:40
*** spenap has quit IRC17:46
*** harbaum has quit IRC17:47
lardmanmore questions :) using a Label, how do I change the background colour?17:59
lardmanor is the background transparent already?18:00
*** BrettQ has joined #harmattan18:00
javispedroso both Nokia and the WebOS use the exact same propietary library to control the vibrator18:01
javispedroImmersion VibeTonz18:01
* javispedro notes in its list of features the ability to set "vibratory ringtones"18:01
lardmanjavispedro: is control of the vibrator so complex that it requires a closed library?18:02
javispedrolardman: I'm having a lot of similar surprises recently18:02
javispedrolardman: yesterday found out that the piece of code that converts raw magnetometer reading into bearing is also closed and licensed from a 3rd party18:02
mikhas_lardman, yes it is complex18:04
lardmanmikhas_: fair enough18:04
SpeedEvilIt's not that complex18:04
SpeedEvilIt's annoying.18:04
SpeedEvilBut it's not actually hard.18:04
SpeedEvilSpherical harmonics of the field are easy to compute.18:05
mikhas_lardman, we find a specially designed feedback and then they need to figure out how to do it18:05
javispedrohey, less work for me. I can just now apply Palm's SDL Haptics patch to my SDL.18:05
SpeedEvilThen you just use that and the derived gravity vector, bit of filtering, and out pops direction18:05
mikhas_Easier to just buy the stuff.18:06
lardmanmikhas_: ah I see, I didn't realise it did haptic feedback too18:06
lardmanI was thinking of tuning for ringtones :)18:06
* javispedro finds the ringtone part more difficult18:06
javispedrobut I guess I don't really value the current state of haptic effects on the N950 (nonexistent)18:07
javispedroonly the vkb seems to produce some18:07
mikhas_With complex I judge based on the amount of issues I've seen.18:07
javispedroSpeedEvil: probably Meego will be looking for an OSS version of such code soon :)18:07
*** BrettQ has left #harmattan18:08
mikhas_javispedro, why?18:08
javispedrorealmeego/meegoCE I mean18:08
javispedroI guess they want all those nifty q{,ml}compass applications to keep working18:08
Stskeepsjavispedro: i think we have a redistributable license18:08
*** BrettQ has joined #harmattan18:09
Stskeepsyou'd be surprised how difficult it is to get open licenses for bleeding edge mobile tech18:11
javispedroeither way, I'm with SpeedEvil, it just requires someone figure out the math and get whatever magnetic model data is required18:11
javispedroaka noone will ever do it :)18:11
lardmanboth of those are described in detail on wikipedia iirc18:12
javispedro"We cover the following: ... Explain how to calculate a bearing from this data."18:15
lardmanI see they correct for declination18:17
*** djszapi_ has joined #harmattan18:17
*** BrettQ has quit IRC18:17
lardmanwould be interesting to compare angle of magnetic field lines with accelerometer18:17
javispedrothat was the reason I was thinking a model would be required18:17
Stskeepsjavispedro: no complaints for any open impl :P18:17
javispedrodepending on the quality of the open impl, data might be of worse quality18:18
lardmanjavispedro: there is a model somewhere I dug it out a few months back when we were talking about using compass field line angles as a location method18:18
SpeedEvilThere is widely available opensource code.18:18
*** BrettQ has joined #harmattan18:18
*** CepiPerez has quit IRC18:19
SpeedEvilFor example - has usable code18:22
*** wazd has quit IRC18:24
* RST38h is kinda back18:24
RST38hAnything new and/or exciting?18:24
javispedrosteve jobs left but I guess you knew that18:25
RST38hEarthquakes? Financial crisis? Black plague? Steve Jobs hitting the bucket?18:25
javispedroalso cmdrtaco leaving slashdot18:25
RST38hjavispedro: Good, let us start our timers18:25
javispedroweird week.18:25
RST38hjavispedro: ~5 years until Apple gets finished18:26
javispedrothe question would be wether that happens before or after Nokia is finished ;)18:26
RST38hPoor CmdrTaco will have to do something morespectacular in order to get my attention though.18:26
RST38hjavispedro: After. Not really a question.18:26
RST38h"The departure of Steve Jobs as boss of Apple was expected to result in a mass sell-off of the overpriced vendor's shares this morning, but the news has been overshadowed by larger events."18:27
*** BrettQ has quit IRC18:27
RST38hjavispedro: But August is certainly a month where fabrique of reality is thin and tends to tear =)18:28
*** rcg has quit IRC18:33
*** frinring has quit IRC18:34
*** BrettQ has joined #harmattan18:38
alteregolcuk: ping18:43
alteregolcuk: cool news from alex. :)18:43
lcukcool and not so cool18:44
lcukalex is a star18:44
*** jkt has quit IRC18:45
lcukalterego, providing I can find footing knowing more about microsoft phones would not be a bad thing18:45
RST38hlcuk: pretty much all you need to know about MS phones is "C#"18:46
alteregoWell, I'd be really interested in doing it and I have a fair amount of C# etc experience.18:47
RST38hwhich automatically kills all high-perf applications18:47
RST38hThere *is* a native SDK, but MS onlyships it to selected developers18:47
alteregolcuk: unfortunately I think I'm out of the country in oct :/18:47
alteregoBut depending when it is I might be able to fly back from the US for it.18:48
djszapi_lcuk: are you experienced with C# ?18:48
*** SpeedEvil has quit IRC18:50
*** SpeedEvil has joined #harmattan18:51
* lardman is sick of QML18:53
lardmantime for some nice C++ code18:54
lardmancan't believe I called C++ nice, but that's the effect QMl has on you18:54
lardmanthough actually QML is growing on me, should make for some nifty looking UIs18:54
RST38hI actually found a way to abstract QML18:55
javispedrois it XML based?18:55
RST38hSo that now I can go into a QML dialog,return from it, and still retain my original QWIdget18:55
RST38hjavispedro: No, CSS+Javascript18:55
javispedroRST38h: as on your apps? But it is slow..18:55
* lardman retreats into the wonders of barcode decoders for some light relief18:56
RST38hThe two really bad problems are 1) QML is so heavy that creating a QML widget takes 3-5 seconds and 2) no way to run QML as part of my event loop, it needs its own18:56
lardmandecoding a full res photo from the N950 takes too much time for my liking18:56
RST38hjavispedro:Well, you have seen how I have done it (packages are out there)18:56
RST38hjavispedro: aside from QML taking 3-5 secs to show the UI, it lookssemi-decently18:57
lardmanapp startup seems quite slow, but the there's that cache thing that I ought to checkout eventually18:57
RST38hlardman: I am afraid that in my case it is not about loading .sos18:57
*** rcg has joined #harmattan18:57
RST38hlardman: Looks like just *creating* the hierarchy, loading all the images, etc takes time18:57
RST38hstarting up JS engine too, probably18:58
mikhas_RST38h, I doubt current state of QML makes it suitable for "complex" UI's18:58
RST38hmikhas: I do not have anything "complex". Just a file selector, a config dialog, and an about dialog18:58
mikhas_erm, oops18:58
lardmanwhat res are the camera photos on the N950?18:59
RST38hWould have SPARQL-based file selector, but the SPARQL QtQ component is not on our devices18:59
javispedroRST38h: oh, see the conversion I had today about that18:59
javispedroRST38h: it is fixed on next fw19:00
RST38hThere are basically two reasons to use QML: 1) Nokia says so and 2) They managed to make all the other UI options (including Qt) barely usable19:00
RST38hjavispedro: cooolio, so I will be able to have a nice file dialog =)19:00
MohammadAGi'm up for fixing the qt theme19:01
*** jkt has joined #harmattan19:01
javispedroMTF is also usable19:01
*** jkt has quit IRC19:01
*** jkt has joined #harmattan19:01
* RST38h makes a move to hug MohammadAG19:01
MohammadAGi could get more done with Qt19:01
javispedroand it is still quite fast, ~1sec for unboosted boot for hello world app19:01
lardmanMohammadAG: +119:01
RST38hjavispedro: Yes, but MTF requires you to rewrite your whole program in some dead tongue19:01
MohammadAGan MTF Qt Designer plugin would be awesome19:01
javispedrothey look all dead to me.19:01
MohammadAGRST38h, QML does the same19:02
RST38hjavispedro: I actually tried using MTF but stopped after finding out that it is totally incompatible with anything19:02
MohammadAGI wrote my Qt/C++ programs since forward compatibility was promised19:02
RST38hMohammadAG: Quim continues saying that QML has a future =)19:02
javispedroit is not incompatible with graphicsview ;P19:02
MohammadAGoh nice19:02
javispedrobut yeah, MTF is NOT Qt19:02
MohammadAGthe bug in fremantle where it thinks music is still running is there19:02
MohammadAGMTF is a KDE wannabe19:02
RST38hQWidget-based stuff and MTF are incompatible19:03
lcukdjszapi_, I know .net stuff, spent years doing ;)19:03
RST38hok, back to the hotel19:03
RST38hlcuk: C# is not VB19:03
RST38hlcuk: C# is Java19:03
RST38hsame vm nowadays though19:03
javispedro though if he has done he knows the ms class names19:03
MohammadAGRST38h, give me one language that's been cross compatible19:04
javispedro/C ;)19:04
djszapi_lcuk: excellent, you need to teach me at some day :)19:04
MohammadAGjavispedro, in the sense that you don't write your own ui19:04
MohammadAGtake my app, sociality as an example19:05
MohammadAGsexy on maemo 519:05
MohammadAGlooks crappy on a desktop, but still somewhat good19:05
MohammadAGsucks on meego and symbian19:05
javispedro(after a long train of thoughts that is too long to explain) this is what happens when you buy a toolkit library.19:07
MohammadAGQML, good concept, but I hate web development cause of the syntax19:07
lcukdjszapi_, lol19:07
javispedroalso: slow.19:08
javispedroWe now have hw where all applications now launch in second-times and we use it to add another layer of abstraction so that we are back into the 5-6 seconds range19:08
MohammadAGI moved from python to C++ cause I hate interpretted code19:08
*** fiferboy has joined #harmattan19:09
MohammadAGcan someone do some binary patching so the damn quick launch opens even when the device isn't top up19:10
javispedroisn't it mostly useless?19:11
javispedrothe shortcuts are not configurable19:11
MohammadAGno, thanks to the absence of a camera key19:11
MohammadAGit's the fastest way to open the camera for me19:12
javispedroMohammadAG: file a device bug19:12
MohammadAGwhat if they say patches welcome?19:13
javispedro"binary" patches welcome?19:13
MohammadAGno, hardware ones to add a camkey :p19:13
MohammadAGjavispedro, it's a design decision19:14
MohammadAGopen the bar and flip the device to landscape mode19:14
MohammadAG(even when the keyboard is open)19:14
javispedroyeah, I've teste19:14
javispedrobut design decision or not, looks like a user confusing bug19:14
MohammadAGanother thing19:15
MohammadAGcan't someone also binary patch the regex that decides which status plugins to load19:15
MohammadAGinstead of volume, ******19:15
javispedroI would prefer to zap the person that wrote that19:16
MohammadAGblame aegis, it's open source19:16
*** arcean has joined #harmattan19:16
*** elzalem has quit IRC19:17
*** lamikr has joined #harmattan19:17
*** elzalem has joined #harmattan19:20
*** fiferboy has quit IRC19:22
*** achipa has quit IRC19:31
MohammadAGjavispedro, query me19:33
*** piggz has joined #harmattan19:41
*** smoku has quit IRC19:50
*** sudanix has quit IRC19:52
*** wazd has joined #harmattan19:58
*** [XeN] has joined #harmattan20:00
DocScrutinizerMEH, on unplugging device in act_dead (aka "not powered on") from charger, it blocks power-on for several seconds as it first needs to finish shutdown from act_dead state. Extremely annoying / confusing20:01
javispedrosame as N81020:01
wazdheya all20:01
MohammadAGyeah, quite annoying20:01
MohammadAGsame happens on power off20:02
faenil_MohammadAG: yeah20:02
Stskeepswazd o/20:02
MohammadAGI use /sbin/reboot20:02
MohammadAGwill editing .bashrc or .profile eff up aegis?20:02
javispedrothe ones in your home, surely not20:03
MohammadAGI want export PATH=$PATH:/sbin:/usr/sbin20:03
DocScrutinizerMohammadAG: for power-off it's at least kinda obvious why it is that way20:03
lardmanhmm, can I not kill -9 something because of aegis?20:03
lcuklardman, nope20:03
MohammadAGdevelsh lardman20:03
DocScrutinizerwhile unplugging from charger you *think* it *is* off20:03
lardmanI am root, ssh20:03
djszapi_lardman: you can if you CAP::kill rights20:03
djszapi_* have20:03
lardmanah ok, so it is aegis related20:04
djszapi_yes, of course.20:04
lardmanI need to devel-su to get the rights20:04
DocScrutinizerand develsh might cure it20:04
DocScrutinizeras ssh root isn't exactly develsh20:04
lardmancool, will try that next time, thanks chaps20:04
djszapi_lardman: nope20:04
DocScrutinizeror devel-su, yeah20:04
MohammadAGdevel-su is just a working su20:05
MohammadAGno aegis privs20:05
djszapi_not true20:05
MohammadAGwell, i still need to use develsh20:05
djszapi_su and devel-su are quite different20:05
MohammadAGsu is broken20:05
MohammadAGthanks to ash20:06
javispedrodamn, time passes way too fast.20:06
MohammadAGwhich reminds me, uni in 2 days :/20:07
djszapi_devel-su would be an unneccesary su duplication without aegis privs20:08
MohammadAGsu doesn't work anyway20:08
MohammadAGneed setuid20:08
djszapi_devel-su is not targetted to duplicate the su theory.20:09
javispedroyou could technically implement a su without setuid using aegis20:09
lardmanhmm, can't kill it even with devel-su20:10
javispedroyou want develsh, not devel-su20:10
djszapi_javispedro: sure, very painfully :)20:10
djszapi_lardman: that is why I said above "nope"20:10
*** elzalem has quit IRC20:10
javispedroDocScrutinizer: use my impromptu signer to fix that20:10
lardmanthanks javispedro20:11
lardmandjszapi_: hard to know what the nope referred to20:11
djszapi_lardman: I answered to you, so probably your last post compared to that20:11
DocScrutinizerjavispedro: first of all it demonstrates that >>ssh -l root N950<< is not the same ID root as the one you get from devel-su20:11
djszapi_lardman: anyway, develsh will probably not work on newer images either20:12
javispedroplease stop confusing deve-su and develsh20:12
DocScrutinizerI.E. has different capabilities from aegis20:12
javispedroit is me who is confusing20:12
javispedrodevel-su = su + develsh20:12
djszapi_su has nothing to do with devel-su20:12
javispedrodjszapi_: yes20:12
djszapi_ok, then yes ;)20:13
javispedrodjszapi_: please recheck as you're probably confusing them like what I was doing moments ago20:13
DocScrutinizerno, develsh actually gives me yet another power than devel-su20:13
javispedroDocScrutinizer: sure20:13
djszapi_entirely sure20:13
MohammadAG20:12 javispedro: devel-su = su + develsh20:13
djszapi_entirely not sure20:13
DocScrutinizerno way20:14
javispedroo really?20:14
djszapi_yes, CAP::kill is not available in develsh20:14
hiemanshuMohammadAG: nope20:14
djszapi_so that is not gonna be a long term memoir20:14
hiemanshudevel-su != su + develsh20:14
MohammadAGdevelsh > devel-su20:15
MohammadAGyou can't kill with devel-su alone20:15
hiemanshuwell no20:15
MohammadAGyou need develsh20:15
hiemanshuyou need to devel-su then develsh20:15
MohammadAGif it's to kill a root app yes20:15
javispedrolet's do this the proper way20:16
MohammadAGyou just need develsh to killall meegotouchhome20:16
npm" This APT has Super Cow Powers. " ... doesn't exactly give one the warm fuzzies one's phone hasn't been hacked :-)20:16
MohammadAGapt-get moo20:16
djszapi_MohammadAG: for the time being, it might work, not in the future though, so I would not get used to it20:17
hiemanshuis down?20:17
MohammadAGget used to what?20:17
javispedrossh user@ < devel-su < develsh20:17
javispedrossh user@ < devel-su <<<<< develsh20:17
djszapi_MohammadAG: that approach, actually develsh credentials were radically confined20:17
hiemanshujavispedro: and just plain develsh?20:18
hiemanshujavispedro: without devel-su20:18
javispedroI didn't knew you could do that20:18
hiemanshuyou can20:18
javispedrossh user@ < devel-su < ssh root@ < devel-sh as user < devel-sh as root20:19
javispedrothere, if you want I can make a QML app with nice arrows and transitions20:19
MohammadAGplease do20:19
hiemanshujavispedro: that would be nice20:19
MohammadAGwith a C++ backend20:20
hiemanshuand one with a JS backend too20:20
hiemanshu'SSO website will be offline for a few hours today while we complete website enhancements.' :(20:20
djszapi_javispedro: what is the password for user ?20:20
javispedrodjszapi_: invalid20:21
javispedroaka it has no password20:21
hiemanshudjszapi_: you can set is, devel-su, develsh, passwd user20:21
infobothiemanshu meant: djszapi_: you can set it, devel-su, develsh, passwd user20:21
MohammadAGuserme? :p20:21
hiemanshuMohammadAG: lol20:21
hiemanshuMohammadAG: able to access RDA?20:21
djszapi_hiemanshu: simple passwd user works.20:21
MohammadAGhiemanshu, i'm on the N95020:21
hiemanshudjszapi_: iirc, I got a permission denied last time I tried20:22
MohammadAGbouncers rock :p20:22
hiemanshuMohammadAG: ah, irc-chatter? :P20:22
MohammadAGhiemanshu, develsh after devel-su then passwd user20:22
hiemanshuMohammadAG: thats what I said20:23
MohammadAGdoes irc-chatter reply to a CTCP version?20:23
hiemanshuIIRC no20:23
* hiemanshu checks20:23
djszapi_Where can I download this busybox-devel-su packge ?20:23
djszapi_* package20:23
hiemanshuMohammadAG: actually it does20:23
djszapi_it is not here.20:23
hiemanshuMohammadAG: Received unknown CTCP IRC by MohammadAG!~MohammadA@Maemo/community/contributor/MohammadAG with arguments: Chatter, the first MeeGo IRC client20:23
MohammadAGapt-get -d --reinstall busybox-devel-su20:24
MohammadAGit'll be in /var/cache/apt/archives20:24
hiemanshuit should reply CTCP VERSION IRC Chatter*20:24
hiemanshuthat should be fixed20:24
djszapi_MohammadAG: it will hardly work if it is not in the repository.20:24
MohammadAGhiemanshu, cool20:24
MohammadAGdjszapi_, it's there afaik, sec20:24
djszapi_MohammadAG: link ? Because what I pasted did not contain it.20:24
hiemanshucan anyone access sso or is down just for me?20:25
MohammadAGit works fine for me djszapi_20:25
djszapi_MohammadAG: yes, because it is probably binary20:25
djszapi_which does not help me without a link..20:25
javispedrodjszapi_: every devkit N950 has  it installed20:26
djszapi_javispedro: I told couple of times, I do not use old software :)20:26
* javispedro is rather worried about that "new" software20:26
djszapi_MohammadAG: could you please publish the package somewhere if you do not know the link ?20:26
djszapi_javispedro: we do not need devel-su since we have aegis-su20:26
djszapi_which is much more powerful20:26
djszapi_so there is no point to have devel-su for us by default...20:27
MohammadAGdjszapi_, does the license allow me to?20:27
javispedroMohammadAG: busybox is GPL20:27
javispedroso it would be a license violation for Nokia20:27
javispedro(if anything)20:28
djszapi_MohammadAG: yep, I could install it anyway by having the same sources.list20:28
DocScrutinizerRM680-22-6_PR_RM680:~# aegis<tab><tab>20:28
DocScrutinizeraegis-constrain        aegis-deb-util         aegis-install-journal  aegis-preinst-hook aegis-deb-add          aegis-developer-mode   aegis-loader           aegis-session20:28
DocScrutinizeraegis-deb-info         aegis-exec             aegis-origin           aegis-unload aegis-deb-release      aegis-fixup            aegis-postinst-hook    aegisfs20:28
DocScrutinizerRM680-22-6_PR_RM680:~# aegis20:28
djszapi_but I do not mix it up with my setup20:28
*** qgil has quit IRC20:28
MohammadAG djszapi_20:29
djszapi_many thanks.20:29
MohammadAGyw :)20:29
djszapi_I always missed the link getter option from dpkg20:29
djszapi_we have such a one for "pacman".20:30
djszapi_my qt package manager also has one ;)20:30
javispedrodebian has dget20:30
javispedrobut it's kind of a helper ontop of dpkg20:30
*** cpscotti has quit IRC20:30
MohammadAGdjszapi_, does your new software have perl installed?20:32
javispedrothey cannot really remove perl20:32
javispedrohalf of debian dies right there if you do it20:32
djszapi_javispedro: nope, your scenario does not work here20:32
MohammadAGperl, not perl-base javispedro20:32
MohammadAGdpkg-repack doesn't work on my N95020:32
djszapi_devel-su has way much fewer capability than root ssh + develsh20:33
djszapi_and actually even different type, sometimes.20:33
MohammadAGas illegal as it may be, i plan on pulling skype from the N9 on RDA20:34
djszapi_that is completely illegal20:34
MohammadAGi know20:34
djszapi_well, I have only perl-base20:35
MohammadAGdistributing it is, copying it for personal use isn't ;)20:35
djszapi_not true20:35
djszapi_you need to get certificate for that from skype, that is what N also does.20:35
MohammadAGyou need a certificate for skyhost20:35
MohammadAGnot account-plugin-skype20:36
DocScrutinizerdjszapi_: could you tell me why /bin/develsh (-rwxr-xr-x    1 root     root          3172 May 17 11:05 /bin/develsh) seems to be a ELF that just calls busybox, rather than a hardlink to the busybox binary?20:36
MohammadAGDocScrutinizer, aegis caps thing20:36
javispedrono setuid bit20:36
DocScrutinizerMohammadAG: would work on a hardlink as well, no?20:36
javispedroprobably uses aegis to su20:37
DocScrutinizerjavispedro: ^^^20:37
MohammadAGDocScrutinizer, isn't aegis tied to the kernel?20:37
javispedroas I mentioned20:37
javispedro<javispedro> you could technically implement a su without setuid using aegis20:37
MohammadAGaka it doesn't matter what the path is20:37
DocScrutinizeraegis-exec might make sense, yeah20:37
*** djszapi_ has quit IRC20:37
javispedroaegis loves hashes, remember. if you give it permission to su to the busybox binary you give su permission to all busybox applets20:37
*** djszapi_ has joined #harmattan20:38
javispedrothus, they made a different binary, and gave it su privilege.20:38
* DocScrutinizer points at >>-rwxr-xr-x 1 root root 3172 May 17 11:05 /bin/develsh20:38
javispedroah sorry.20:39
javispedroI was confusing develsh with devel-su again20:39
* javispedro 's brain elsewhere and should stop20:39
djszapi_javispedro: do you understand what devel-su is for ?20:39
djszapi_maybe if it is clear you could distinguish them better :)20:39
DocScrutinizerI'd like to decompile develsh, but I think an strace shall do20:40
javispedroeither way, my point was valid, the same happens to devel-su:20:40
djszapi_why would you do that ?20:40
javispedro$ ls -l /bin/devel-su20:40
javispedro-rwxr-xr-x    1 root     root         28248 May 18 18:06 /bin/devel-su20:40
DocScrutinizerwooohoo 10 times the size20:40
javispedrocause devel-su is a copy of busybox20:40
djszapi_yes, of course everything is like that in /bin/20:40
DocScrutinizerduh what?20:41
javispedroa copy that is probably trimmed to only do "su"20:41
DocScrutinizerbusybox has 28k size????20:41
djszapi_javispedro: I think it might help if you can get to know devel-su and what is it for ;)20:41
javispedroDocScrutinizer: I do not have bash, but if you do "exec -a busybox devel-su" from it you'lls urely get a busybox help message20:42
DocScrutinizerI gather devel-su is for HARM what sudo gainroot was for fremantle20:42
djszapi_it is tied to aegis, it does not try to reinvent what aegis was designed against.20:42
javispedroimho, devel-su = a special build of BB that only does "su" that has a specific aegis token that allows it to su without having the setuid bit.20:43
javispedrodevelsh = a binary that execs() /bin/sh but has a bunch of aegis tokens granted (a large bunch)20:43
djszapi_I must say, it is about correct20:44
DocScrutinizeranyway that'S what develsh does:
djszapi_of course, devel-su is more than just a setuid20:45
javispedroDocScrutinizer: aka execve(/bin/sh)20:45
djszapi_but that is one of the main purpoess.20:45
DocScrutinizerjavispedro: that's what I thought as well, and you could do the same aegis magic to a hardlink to busybox binary aiui20:45
javispedrothat's what not I'm sure about.20:46
javispedroaegis probably considers hardlinks the same file20:46
djszapi_aegisfs does not handle hard links imo20:46
*** harbaum has joined #harmattan20:47
DocScrutinizerwe're back to my query session of a few days ago, where I asked how a record in aegis' database is linked to a certain binary20:47
djszapi_what do you mean by "record" ?20:48
DocScrutinizerI assumed it was via the fully qualified pathname20:48
DocScrutinizer+ inode20:48
DocScrutinizerwell, a records that has all the credentials that develsh gets granted20:48
djszapi_sorry, it might be my missing technical knowledge, but I do not understand the question. Could you please ellaborate on this ?20:48
DocScrutinizerwhen I start develsh, *something* looks into aegis internal database (or kernel's image thereof, or whatever) and finds a set of data applicable to that binary20:49
djszapi_DocScrutinizer: what is the question, I am still not getting it.20:50
javispedrothis vibrator library is such a trade-secret it even requires me to submit a license key to it in the initialize function.20:50
DocScrutinizerthis set of data (i called it "record") says >>give develsh cap::chroot<< and whatnot else20:50
javispedroDocScrutinizer: "something" is the kernel itself20:50
djszapi_oh Doc, I told you that.20:50
djszapi_what I said is about: the restok.conf is not parsed by the daemon20:51
djszapi_by a daemon*20:51
djszapi_validator init maps it into the kernel20:51
Stskeepsjavispedro: immersion is fairly properitary yes20:51
djszapi_and when a new package gets installed, dpkg updates it.20:51
DocScrutinizerwhat is the primary key in this database, I.E. what makes develsh develsh? will it be develsh still when it moves from /bin/ to /sbin/?20:51
djszapi_DocScrutinizer: nope20:51
javispedroStskeeps: fortunately, seems that "The device already has a valid license embedded which replaces the dummy value. You only need to remember to call the API with the dummy license using the SetDeviceProperty method."20:51
djszapi_DocScrutinizer: just look at how the aegis manifest works, you sets the rights for absolute paths20:52
javispedroDocScrutinizer: it's kinda hard to explain20:52
javispedroDocScrutinizer: between reboots the DB is indexed by path name20:52
DocScrutinizerdjszapi_: I didn't even mention "daemon" - what I said was >>*something* looks into aegis internal database (****or kernel's image thereof****, or whatever)<<20:52
javispedroDocScrutinizer: once the system is up, it is indexed by device number + inode number20:52
Stskeepsman, aegis is really the most mentioned word in here20:53
javispedroDocScrutinizer: the validator-init process during early userspace reads pathnames from a file and passes device#+inode# to kernel20:53
DocScrutinizerdjszapi_: that's what I always suspected and you said "no, it's not working this way"20:53
djszapi_DocScrutinizer: really ?20:53
tommahmm... pkgmgr list installed is bugging...20:53
djszapi_DocScrutinizer: strange, because I explained these things yesterday or the day before yesterday20:53
DocScrutinizeryes, I even think I mentioned fully qualified pathname several times20:54
javispedrothus, it's not fully pathname indexed, but it's not fully inode indexed either :)20:54
*** deimos has joined #harmattan20:54
djszapi_DocScrutinizer: I am not sure about this inode things, to be honest20:55
djszapi_but the path is full path, completely.20:55
djszapi_as in absolute.20:55
DocScrutinizerjavispedro: HAH, so it's linked to the inode+device uple in runtime, while in database it's pathname20:55
javispedrotherefore, if /a and /b point to inode 3 (assume busybox), both will be the same from Aegis PoV.20:55
javispedrodespite the fact that you will have two records in the pathname db.20:56
DocScrutinizerBWAH, sure20:56
DocScrutinizergot it20:56
DocScrutinizermuhahahaha really funny20:56
javispedroinode-based vs path-based is a rather large discussion in the linux security modules community20:57
DocScrutinizerhey, I think I can guess why20:57
javispedroiirc, selinux is inode-based while apparmor is path-based20:57
DocScrutinizerI'd have thought kernel runtime uses pathname PLUS inode20:58
djszapi_let us clarify it tomorrow for sure...20:58
DocScrutinizerprolly overkill20:58
DocScrutinizerdjszapi_: sure20:58
DocScrutinizerthanks djszapi_20:58
javispedroruntime actually uses: device number, inode number, file uid, file gid, perms, and hash.20:58
*** vladest has joined #harmattan20:58
javispedro(only dev# and inode# for indexing though)20:59
javispedrobut if any of the above changes, binary is considered different.20:59
javispedroand thus any associated tokens won't be granted (and depending on enforcement, -EPERM, but you know that already :D )20:59
DocScrutinizeryeah, and if it's a mandatory file, you earned a "Please Reflash!" MALF :-D ?21:00
djszapi_javispedro: the interesting part is not the token :D21:00
djszapi_but what DocScrutinizer said.21:00
djszapi_but unfortunately, I do not know this bit till tomorrow..21:00
djszapi_DocScrutinizer: not that simple, but saying very simply: yes21:01
DocScrutinizernewsflash, bbl21:01
djszapi_iirc smack is also "inode based", but at any rate, let us clarify it from authorized source tomorrow :)21:02
djszapi_DocScrutinizer javispedro: do you think it means any security hole ?21:03
djszapi_because I do not see that way either.21:03
djszapi_so I did not get the "muhahahaha really funny" bit.21:03
DocScrutinizerdon't think it's a security hole21:04
DocScrutinizeralas :-D21:04
javispedroin fact i remember the classical common criticism for apparmor was that21:04
javispedroln /etc/shadow /home/user/myfile was all you needed to be able to read the passwords ;)21:05
javispedrocause, being path-based, /home/user/myfile was a different file ;)21:05
DocScrutinizerthe funny bit was about the detail that hardlinks don't work here, and why21:05
djszapi_but this is one difference between smack and aegis.21:06
DocScrutinizercould get you rather confused if you don't know about the internals and think "duh, I don't need /bin/develsh size 3297, I can do this with a hardlink" and then find you nuked your system21:06
djszapi_all in all, they will be abstracted out by libcreds2 though.21:06
djszapi_or libcreds3, even.21:06
*** rm_you| has quit IRC21:07
djszapi_DocScrutinizer: does fuse support hard links ?21:08
DocScrutinizerfrankly dunno21:08
djszapi_mmh, it seems it does21:08
djszapi_it might be that we had just issues with those in aegisfs and fixed by now21:08
DocScrutinizerI was already temped to claim inode based is probably faster, but that's really not exactly a criterion here21:10
djszapi_hahaha aegisfs codebase seems to be very secury21:11
djszapi_it is not on gitorious21:11
DocScrutinizerfor sure inode based is simpler to evaluate for security holes. Pathnames are really complicated with lots of dark corners21:11
DocScrutinizersee bindmounts, other stuff21:12
DocScrutinizerwhatnot else21:12
djszapi_DocScrutinizer: so when will you ship the USB host mode ? ;)21:13
DocScrutinizerbut then otoh it's harder to debug, as you basically can't get the pathname for an inode21:14
*** javispedro has quit IRC21:14
DocScrutinizerdjszapi_: as this is no security hole, I still have the problem of missing privileges in developer mode, and missing possibility to add kernel modules for stock kernel. So probably hostmode is not feasible21:15
djszapi_well, you had to make an own kernel for N900 as well, anyways21:16
DocScrutinizersure, but that kernel didn't deprive users unstalling it from a good chunk of the more important apps. On harmattan it's still unclear if we can run own kernels at all, and it's even more unclear and highly doubtful if a normal harmattan system incl all apps like dialer, maps, drive, could run on such a new kernel21:18
infobotDocScrutinizer meant: sure, but that kernel didn't deprive users installing it from a good chunk of the more important apps. On harmattan it's still unclear if we can run own kernels at all, and it's even more unclear and highly doubtful if a normal harmattan system incl all a...21:18
djszapi_why not, you just doom a certain functionality in aegis validator in my understanding21:18
djszapi_you do not kill out the runtime policy framework (you do not actually even touch that)21:19
DocScrutinizerI think aegi't hack it that way21:19
DocScrutinizerI think aegis is too good for that, you can't hack it that way21:20
djszapi_I do not think that is not possible, it is just like to open up the possibilities about loading a module into the kernel or extend it21:20
DocScrutinizerif maps depends on Nokia Account, and Nokia account depends on working aegis, then I think we will not see maps running on a custom kernel21:20
djszapi_so in theory, I do not see any issue why you could not modify aegis according to that only requirement21:21
djszapi_why not ?21:21
DocScrutinizerif aegis is as cleanly designed as I guess it is, it simply will cease to work if you "hack" anything of aegis21:21
DocScrutinizerit's a chain of trust, and editing any arbitrary bit causes "open mode"21:22
djszapi_come on you modify the aegis design according to your design :D21:22
djszapi_it is not a saint design that cannot be modified.21:22
djszapi_thing is that nobody tried it out yet (what I said).21:23
djszapi_that could be the first step and then let us see whether it is enough or not21:23
DocScrutinizerloading a kernel that's not properly hashed and signed causes TPM in OMAP to lock, and from then on you need to powercycle to ever again access the root cert21:23
djszapi_but you can surely extend validator whitelist with your own kernel moudle.21:23
*** vladest has quit IRC21:23
*** vladest has joined #harmattan21:24
djszapi_or just drop the kernel signing...21:24
*** sudanix has joined #harmattan21:25
DocScrutinizerdjszapi_: I know a friggin lot about TC basics, and how they work together to forbid those things you suggested. I'm both confident the aegis devels were smart enough to follow this well known concept, and also I'm not willing to play the pen-tester for aegis21:25
*** djszapi__ has joined #harmattan21:26
*** djszapi_ has quit IRC21:26
DocScrutinizerI'm willing to start hostmode development as soon as I get instructions how to install and run a modified kernel and a semi-working HARM system on top of it.21:27
DocScrutinizerI'm not developing the methods to get such a custom kernel and system running21:27
DocScrutinizeras that's not what I offered to do21:27
DocScrutinizeralso I need a system where a random binary like i2cscan runs without aegis bitching at me about missing permissions to access I2C-bus etc21:28
djszapi__what is wrong precisely about packaging that ?21:29
DocScrutinizerthis is a reqirement strictly for development, not related to later user package21:29
*** sudanix_ has joined #harmattan21:29
djszapi__yes, but does it require rapidly changing credentials during the development phrase ?21:30
*** sudanix has quit IRC21:31
DocScrutinizerdjszapi__: well, maybe packaging even might work, just it's too steep a startup hurdle to build each silly tool like e.g. bash or i2c-tools myself, I'm loosing interest when I can't access a certain pool of tools I expect to see on each decent development platform21:31
djszapi__well, you can always run relaxed stuff21:31
djszapi__but these trivial tools are surely not hard or too much time to package.21:32
djszapi__someone can even do for you, I guess.21:32
DocScrutinizerdoesn't work, that's what I said. Maybe I was wrong as I didn't know enough about how aegis works exactly21:32
djszapi__yep, it should work21:33
DocScrutinizerI found out femantle packages install just fine usually, and installed i2c-tools, and of course it worked but wasn't allowed to access i2c-bus21:33
djszapi__just run it in develsh with relaxed stuff21:34
DocScrutinizerI'll give it a second try eventually21:34
DocScrutinizer(well, a 5th or 9th, to be precise :-D )21:34
djszapi__btw, you did not hear this relaxed stuff from me, right ? ;)21:35
DocScrutinizersee, I'm always afraid I get another "please reflash!" and there's NO way to backup all the stuff already installed and configured21:35
djszapi__yep, that is scary and the forever disadvantage of any old image of any sort.21:36
djszapi__so Intel might drop MeeGo, really ?21:38
Clintand switch to WebOS?21:39
djszapi__no, they have been playing Android iirc for a while. Trying to port it to some intel architecture.21:39
djszapi__maybe since last december or so.21:39
DocScrutinizerwell, *NOKIA* could easily "fix" this by simply deploying a signed package with a /etc/init/check_malf.conf that does NOT freeze boot when this notorious evil requester pops up21:39
djszapi__DocScrutinizer: that is not that simple21:40
djszapi__if libbb5 damages, no way your system is worthwhile using.21:40
djszapi__just a random example.21:40
DocScrutinizerstill I'm more afraid of an accidental "echo WTF >/var/malf" rather than any fsckup of libbb521:41
Stskeepsgently make love to aegis(TM)21:42
DocScrutinizerand I could just "rm /var/malf" rather than reflash, if boot wouldn't block on /etc/init/check_malf.conf and this evil requester screen21:42
djszapi__Stskeeps: malf management has nothing to do with aegis21:42
djszapi__what they do after a syserror, that is a different team21:43
djszapi__called system-software/services21:43
djszapi__again, just a lack of misunderstanding :)21:43
DocScrutinizerjust it's been aegis that triggered my last syserror :-P21:44
djszapi__anything can seriously...21:44
DocScrutinizeryeah even I can, see  "echo WTF >/var/malf"21:44
djszapi__and aegis is not responsible for something that other team turns into unusable state.21:44
DocScrutinizerit's been aegis-exec after all that triggered that supposed bug that nuked my device21:45
djszapi__that bug is highly unreproducible21:45
djszapi__on the later images.21:45
DocScrutinizerso AMLF handling is rogue, but aegis triggering syserror on random events21:45
djszapi__nope, aegis is not triggering ;)21:46
DocScrutinizerI got no later image21:46
djszapi__that was a plain implementation bug21:46
djszapi__it is not architecture design, so it is not really something serious21:46
djszapi__mainly if it is fixed21:46
djszapi__and actually I am not sure that was an aegis bug21:46
djszapi__that is just one thing it happened by using aegis21:46
DocScrutinizerI got what I got, hw- and image-wise21:46
djszapi__it is really not a strong prove as such21:47
djszapi__"echo WTF >/var/malf" -> I can remove /var/malf :)21:47
djszapi__after touching it ;)21:47
DocScrutinizerbottom line, I'm not excited to try and investigate something on a system where I have to do 2 days of preparations for a test that takes me 30min on fremantle21:48
DocScrutinizersure, but boot the system without removing it :-P21:49
djszapi__cannot you use minicom to remove it ?21:49
djszapi__you do not basically get a prompt at all ?21:49
DocScrutinizerwhen it doesn't boot?21:49
DocScrutinizerno, it does NOT boot21:49
djszapi__so it is not a Ui issue21:49
DocScrutinizer(except into act_dead)21:49
DocScrutinizerdjszapi__: >> less /etc/init/check_malf.conf <<21:50
DocScrutinizereasy to understand21:50
DocScrutinizerrm /etc/init/check_malf.conf and tell me I can do this too, and I'm happy21:50
DocScrutinizerI don't dare to even try to edit this file21:51
djszapi__DocScrutinizer: that file is mentioning the RD certificate for you ?21:51
DocScrutinizererr, dunno, maybe21:52
DocScrutinizermoment please21:52
djszapi__no, nobody can remove from /etc/init/21:52
djszapi__as I said from the beginning, that is integrity protected :)21:52
djszapi__no way, you can remove it21:52
djszapi__except from the dedicated interface subfolders.21:53
DocScrutinizer /usr/sbin/show_malf `cat /var/malf`  probably could check if there's an R&D cert, and show a "DISMISS&BOOT" button if it's a R&D device21:53
djszapi__right, it seems my version is different21:54
djszapi__anyway, let me try this malf thingy with my n95021:54
djszapi__not the loaned one, but the newest version one :)21:54
djszapi__or well, even better with the newest N9 tomorrow.21:55
DocScrutinizerthere's also some /usr/sbin/enter_malf21:56
djszapi__I guess you cannot mount it up from your laptop in malf state21:56
djszapi__and remove that file ?21:56
DocScrutinizerI didn't dare to enter_malf --help :-D21:56
djszapi__I have just done ;)21:56
DocScrutinizerdjszapi__: for sure not21:56
djszapi__well, it is a bug if enter_malf -h/--help explodes, don't you think ? :)21:57
djszapi__Iwould say a very serious one21:57
DocScrutinizeryes, but it's also a bug when aegis-exec -a CAP::SUID bash enters MALF :-D21:57
djszapi__yep, but it is worth to catch them right now and fix it, then never fix it.21:58
DocScrutinizerso you *got* that bug?21:58
DocScrutinizerwhat was the result of enter_malf --help??21:59
DocScrutinizerdid it actually explode for you?21:59
djszapi__/usr/sbin/enter_malf --help21:59
djszapi__/usr/sbin/enter_malf: what do you mean21:59
DocScrutinizerok, I feel venturous22:00
DocScrutinizer(though I should make sure no hammers etc in vicinity ;-P )22:00
DocScrutinizerLOL, what do you mean?22:01
DocScrutinizercould be worse, it could show that notorious requester after instant reboot, with "What Do You Mean? \n Please Try To Reflash!"22:02
djszapi__DocScrutinizer: well, echo WTF >/var/malf put me into malf state, but I could reboot22:02
djszapi__but I guess it is still not nice22:02
DocScrutinizerthought as much, your device probably recovers22:02
djszapi__the Ui is funky after the reboot though :)22:03
DocScrutinizerplease elaborate on "but I could reboot" - did it need two reboots where first boot showed the MALF requester?22:04
djszapi__nop, just one, but this is not the point.22:04
djszapi__I am a bit afraid, if anybody can just trigger it this easily.22:05
DocScrutinizerhmm seems your device does some nasty recovery action?22:05
djszapi__no clue :)22:05
DocScrutinizeryup, THIS *IS* a security issue22:05
djszapi__indeed, will test it with later images.22:06
DocScrutinizerveritable DoS attack22:06
djszapi__of course without R&D cert22:06
DocScrutinizerI already suggested I build an app out of it ;-P22:07
DocScrutinizerthe Nuke-me app22:07
djszapi__what you should really not forget, it is still the best phone OS out there ever. Have you ever tried meego previously, not just de, ce and others ? :)22:08
djszapi__at least from Nokia.22:08
DocScrutinizerdescription: "locks your device on the click of an icon, so NOBODY can ever sniff your sekrit private data"22:08
DocScrutinizer(continued:) "give the app a teasing name, like MyChildPorn, and be sure any cop will click on it first"22:09
djszapi__right as I guessed, enter_malf is also coming along from the sys-soft guys22:10
djszapi__probably that should be the interface22:10
djszapi__but not writing any malf file manually.22:11
djszapi__I would actually even expect that you should have some extra credential to run that binary.22:11
DocScrutinizerwell, I'm not sure normal user can create files in /var22:11
djszapi__so not that like, we defend the phone against the superroot and so forth, and it can brick the device easily.22:11
djszapi__does not matter, user/root/whatnot22:12
DocScrutinizerexactly. That's been the moment when I said "NUFF!"22:12
DocScrutinizerit feels like i'm the stupid devel that gets punished by "Please reflash" for doing sth silly22:13
djszapi__yep, this is really weird22:14
djszapi__but it has nothing to do with aegis as such22:14
djszapi__it is the misuse of aegis from the sys-soft guys22:14
DocScrutinizerand that's been the mood where I packed and stowed away that device22:14
djszapi__or the not-use :D22:14
DocScrutinizerwell, the fact that I can't fix it now, despite I'd know how to fix it, that's:
DocScrutinizerand that IS aegis22:16
djszapi__if you can fix it, you can do anything22:17
djszapi__so no point in security anymore, if you can do anything22:17
djszapi__so it is rather controversial thesis of yours :)22:17
*** harbaum has quit IRC22:18
djszapi__and ofc no sec is no go from usr pov.22:18
*** djszapi__ has left #harmattan22:18
DocScrutinizeryeah, we don't need to iterate this philosophical debate again. I'm not a friend of security by Nokia's courtesy and grace22:20
DocScrutinizerI'm willing to care about security myself22:20
DocScrutinizerand buy me a way to fix things like these myself this way22:21
DocScrutinizermight be related to my age, you know ;-)22:21
jktso, I've tried using the Nokia's one click flasher on my N950, and now it appears that the device won't boot anymore22:22
DocScrutinizerI already learnt to care for myself22:22
DocScrutinizerjkt: reflash ;-)22:22
jktwhen I press the power button, the status LED shines, I hear the vibration and the LCD lights up and shows the Nokia logo22:23
*** sivang has joined #harmattan22:24
DocScrutinizerjkt: and maybe make sure your battery is fully charged prior to that (sorry no, I have no idea how to do that. Except maybe... wasn't there a flasher --option --charge-battery or sth?)22:24
DocScrutinizerooh, so odds are you could actually charge battery then? great!22:24
*** djszapi__ has joined #harmattan22:24
*** jkt_ has joined #harmattan22:24
jkt_and it looks like my IRC box died, too.22:25
jktDocScrutinizer: I've just flashed that device; shall I really do it again?22:25
DocScrutinizerfirst check if it looks like charging when you plug the powered-off device to wallcharger22:26
DocScrutinizer(assuming you have a Nokia wallcharger)22:26
DocScrutinizerif the indicator LED starts flashing slowly, I'd let it sit for several hours, then reflash22:27
jktDocScrutinizer: the flasher tells me that it's at 97%22:27
DocScrutinizerthe flasher tells that? :-o DUH22:28
DocScrutinizerthen it should be OK to reflash22:28
DocScrutinizerjust notice that flashing actually takes almost an hour - allegedly. (didn't stand by and stopped the time, anyway it was loooong)22:29
sivangjkt: how did you come to having to reflash the device? :)22:29
sivangand it is easier like with Maemo?22:29
jktsivang: well. I hit, and there was an advice that one shall reflash22:30
DocScrutinizerwell, more like a suspicion that a reflash might help22:30
sivangjkt: so this is where you report Harmattan bugs! nice!22:31
sivangjkt: thanks for your help I was looking for that.22:31
jktI've tried to "reset to factory settings", but it didn't help, so I tried to "clear the device". At that point, it stopped responding, completely.22:32
sivangjkt: I had this problem, but a couple of reboots and one OVI update fixed that btw.22:32
sivangjkt: how is it not a prototype device btw?22:32
sivangjkt: even though you got it from Ronan22:32
sivangjkt: I have "developer edition, not for sale" on the back22:33
sivanganyway be back later gotta workrave22:33
jktsivang: sorry?22:33
*** smoku has joined #harmattan22:52
*** mzanetti_work has quit IRC23:04
*** javispedro has joined #harmattan23:25
lardmanah, that was remarkably frustrating23:26
lardmanannoyingly it seems that the QFileInfo for a photo changes at some point after it's written to disk23:27
lardmanI wonder if the geolocation stuff opens it and writes in some exif data23:27
lardmanin any case it was messing up my folder watching for photoanalyser23:27
lardmanmacro mode isn't much cop on the N950 I notice23:31
jktso, let's hope someone on the forums has an advice --
*** wazd has quit IRC23:44

Generated by 2.15.1 by Marius Gedminas - find it at!