These are chat archives for abranson/rockpool

23rd
Jul 2016
ruff
@rufferson
Jul 23 2016 06:25
I'm using The Day Today - guaranteed pin delivery with no intervention and minimal interaction (just select subscription topics in settings)
Andrew Branson
@abranson
Jul 23 2016 07:56
Cool will give that a try. The app sync with pebble is cool too!
Watchapp deletion isn't working for me though, rockpoold crashes. will get a log later on today.
if it's functionally working though, i might go for a beta release. i did get informed of a memory leak recently though - have you found any of those lately? someone's oom killer was killing everything because rockpoold had got so large.
Andrew Branson
@abranson
Jul 23 2016 08:41
i've sent a screenshot of a day today pin to the ubupebblers room. hopefully they might be inspired to attempt to merge some of this stuff in.
ruff
@rufferson
Jul 23 2016 10:01
Indeed crashes on deletion, something messed up with locker resync
ruff
@rufferson
Jul 23 2016 10:49
re memory leak - i found several places where items were stockpiled without cleanup. Mostly it's about notifications - old notification engine was adding notifications with no cleanup, and platform similarly adding cleaning up only on notification-closure. which as far as i can see is not always happening. timline is rewritten and new engine is tracking life cycle properly. platform part is still not tracked explicitly, although timeline engine makes sure to send RemoveNotification to the platform whenever it cleans up its pin counterpart.
although older rockpool was crashing on every call end, so i can imagine it going oom for device with no calls ant lots of notifications. eg tablet?
Andrew Branson
@abranson
Jul 23 2016 11:26
someone from jolla sent me a journal dump of it, i think it was a phone
but it was great to see some core people using rockpool :D
another thing i've been told about is the phone hanging after a call, this one: http://talk.maemo.org/showpost.php?p=1510418&postcount=810
ruff
@rufferson
Jul 23 2016 11:30
I personally found several times that my battery was drained by caller-ui which was consuming smth like 50-70% cpu, restarting it helped to calm down cpu
Andrew Branson
@abranson
Jul 23 2016 11:30
all i got from michael on ubupebblers was a 'sure'. so enthusiastic about all the work you've done :)
ruff
@rufferson
Jul 23 2016 11:31
but again - rockpool only sends dbus calls in this respect, it does not intervene into ofono or some lower level, so if caller-ui/manager hangs - it's something bad in dbus handler there
Andrew Branson
@abranson
Jul 23 2016 11:31
i think he really thought sailfish devs would have come over to bazaar/launchpad and re-added the sfos parts. even after he cut them out. ah well.
ruff
@rufferson
Jul 23 2016 11:32
I see they don't struggle much with limited functionality, maybe their users do, but... who cares :)
Andrew Branson
@abranson
Jul 23 2016 11:32
i think we need a video or something of all this new functionality. it's so impressive, it deserves to get some publicity
ruff
@rufferson
Jul 23 2016 11:35
basically once i finish sendText and Weather - we're on par with official pebble app with the only missing functionality being cloud services - health and voice.
with voice I have a plan to use IBM Watson as i said - at least as one of the possible options. And health... I'm not sure i would be able to reverse engineer all that flow of data it dumps
there's still a lot of work to make some companion-api - but that needs some wider audience to brainstorm it
Andrew Branson
@abranson
Jul 23 2016 11:41
the one thing michael's worked on all this time is the health stuff. he's done graphs etc. the ui stuff must be of little use to us, but he spent quite a lot of time trying to unpick the data coming from the watch.
katharine's apologised though, because there's very little documentation of it, and it wasn't in libpebble so it's been hard. better wait for it to turn up in there.
ruff
@rufferson
Jul 23 2016 11:43
ah, so they are doing something after all :) bcz I've been checking from time to time bazaar and the only commits i saw was automatic translation
Andrew Branson
@abranson
Jul 23 2016 11:45
quite, so it should be quite easy for them to merge everything we've done
ruff
@rufferson
Jul 23 2016 11:46
then it could become a good collaboration in the end, if they pickup health (which for some people is main reason to have such watch in the first place) and we do the rest of bells and whistles
Andrew Branson
@abranson
Jul 23 2016 11:46
they'll have to rewrite their platform interface and ui, and i bet there's some mismatch of data elsewhere
but i think he's basically waiting for us to offer him merge requests
ruff
@rufferson
Jul 23 2016 11:47
but their platform is very straight forward, i can migrate it myelf to new engine
Andrew Branson
@abranson
Jul 23 2016 11:47
apparently it'll run on ubuntu desktop too
ruff
@rufferson
Jul 23 2016 11:47
but only with blind port - as i don't have any ubuntu device
Andrew Branson
@abranson
Jul 23 2016 11:47
would you want to though? :)
ruff
@rufferson
Jul 23 2016 11:48
if it could speedup merge - i can spend couple of days. although it will always be a pain to sync two different repos
Andrew Branson
@abranson
Jul 23 2016 11:48
tbh, they took over the pebbled stuff, moved it out of github, didn't tell anyone, then dropped the sfos support completely.
rockpool for ubuntu you mean ;)
ruff
@rufferson
Jul 23 2016 11:49
uhm... yes, that's another way to do it
Andrew Branson
@abranson
Jul 23 2016 11:49
i don't think it was ever reasonable to expect sf/ut collaborations to be hosted in bzr on launchpad
their translation system isn't even compatible
(afaik)
ruff
@rufferson
Jul 23 2016 11:49
and launchpad can happily live with git
Andrew Branson
@abranson
Jul 23 2016 11:49
it's the infra. it's aimed at ubuntu
it's like us hosting on merproject.org :)
ruff
@rufferson
Jul 23 2016 11:51
yeah, but even merproject is pure git, there's just own build system
but bazaar is just another of those NiH-driven absurdities
Andrew Branson
@abranson
Jul 23 2016 11:52
hehe
ruff
@rufferson
Jul 23 2016 11:52
mir is another one %)
Andrew Branson
@abranson
Jul 23 2016 11:53
they're replacing stuff all the time
not sure what their translation system is, but the sfos one is the qt standard one isn't it?
ruff
@rufferson
Jul 23 2016 11:54
sfos is using Qt yes, and they are using gettext - which is lower level and more widely used actually
Andrew Branson
@abranson
Jul 23 2016 11:54
ah ok. all those .pot files?
ruff
@rufferson
Jul 23 2016 11:55
yep. we still have some remnants of gettext in the daemon part, but apparently they are unsused since we're not providing any po files
I was planning to replace them with Qt's tr to have common translation for app and service
Andrew Branson
@abranson
Jul 23 2016 11:55
yep
but i suppose that would break compat with UT
weird they don't use the Qt one - they support more Qt than SF does
might just be michael's preference though
ruff
@rufferson
Jul 23 2016 11:56
i think it's just bcz bazaar supports autotranslation for po files but not for ts
Andrew Branson
@abranson
Jul 23 2016 11:56
of course
it's a shame it lost the nice settings mechanism that was in pebbled too. the UI in that was just emitting config changes to the daemon without needed the dbus
ruff
@rufferson
Jul 23 2016 11:57
i'll leave gettext as it is till we're confident we're not going to merge
Andrew Branson
@abranson
Jul 23 2016 11:58
(i think)
ruff
@rufferson
Jul 23 2016 12:00
well there's slight overhead doing it purely via dbus, on the other hand there's full isolation of the client from service - i.e. client cannot break the service if you write dbus handlers safely
Andrew Branson
@abranson
Jul 23 2016 12:01
yeah, though it's only transmitting config changes, which you can also handle safely
ruff
@rufferson
Jul 23 2016 12:02
but you can accidently break the config making it unparsable and hence losing all the functionality. while to clean up config over dbus you need to send explicit clears to each and every set* call
Andrew Branson
@abranson
Jul 23 2016 12:03
well it was a bit more structured than that, but i see your point
but one of my aims is to have schema validated config everywhere... :)
ruff
@rufferson
Jul 23 2016 12:04
also there's not much functionality linked to the config itself, most of the gui interaction is about runtime
and if you want to offload to the client such elements as app installation and pin management - you then need much higher level of synchronisation - flocks, semaphores, etc.
i do agree though that appdownloader is not the best idea having it in the service
ruff
@rufferson
Jul 23 2016 12:10
should suffice to fetch the file using client's appstore client and then call standard sideload process. Similar with locker - it's sarther appstore part than service. In short - we have a lot of space to improve the design and architecture of the app :)
ah i remember why i put locker to service :) bcz i didn't want to path auth tocken back and forth between service and gui, and service needs token for timeline
*to pass
Andrew Branson
@abranson
Jul 23 2016 12:44
aha I see
ruff
@rufferson
Jul 23 2016 12:50
I'm a bit in the middle of sendText/Weather creation so I'll commit app removal fix with pieces of that stuff
ruff
@rufferson
Jul 23 2016 13:13
or actuallly I'll just stash and switch to master
Andrew Branson
@abranson
Jul 23 2016 13:27
bless git :)
except for force push. i hate force pushing.
had to do my first force push into a jolla repo the other day though. palms were sweaty.
ruff
@rufferson
Jul 23 2016 13:29
I'm quite used to it, as long as you're sure no-one tracking your head - it's quite fine
frequently squashing fixups
Andrew Branson
@abranson
Jul 23 2016 13:31
yeah exactly for squashing, but it's completely against the idea of version control
and the only undo is to ask someone else who might have another copy of the repo, to force push it back :)
ruff
@rufferson
Jul 23 2016 13:33
yes but polluting the history of the big project with small fixups which is discovered immediately after commit - is also a bit silly. reading the log or dissecting then becomes a nightmare
the branch is a remedy %)
Andrew Branson
@abranson
Jul 23 2016 13:34
sure, but that doesn't mean there shouldn't be some sort of potential for an undo somewhere. like the orig_head thing.
some sort of temporary echo of the old state
ruff
@rufferson
Jul 23 2016 13:34
keep the WIP with full track and squash to master once all is sorted
that's what i do, commit all over the road to wip, and before merging to master do the squash and pull rebase to master
although my master is still kinda wip relating to your :)
Andrew Branson
@abranson
Jul 23 2016 21:57
ok there is something odd going on here. i've built it twice more in 'release', and i've got the [D] main:57 - Manifest "/opt/sdk/rockpool/usr/share/rockpool/jsm/RockpoolJSComponents.manifest" modified "" error again
trying a clean
ruff
@rufferson
Jul 23 2016 21:58
you can check during the build whether -DROCKPOOL_DATA_PATH=\"/opt/sdk/rockpool/usr/share/rockpool/ is appearing in the compiler output
it should appear only in debug configuration - as per this stanza in rockwork.pro
CONFIG(debug, debug|release) {
DEFINES += 'ROCKPOOL_DATA_PATH=\"/opt/sdk/rockpool/usr/share/rockpool/\"'
}
and then when defined it overrides this default in main.c

ifndef ROCKPOOL_DATA_PATH

define ROCKPOOL_DATA_PATH "/usr/share/rockpool/"

endif //ROCKPOOL_DATA_PATH

ops
Andrew Branson
@abranson
Jul 23 2016 22:09
More strangeness. My url handler's being replaced with a list of every Android app I have installed. wtf! :)
make[1]: Entering directory `/home/src1/buildwin/rockpool-MerSDK_SailfishOS_armv7hl-Release/rockwork'
g++ -c -pipe -O2 -g -pipe -Wall -Wp,-D_FORTIFY_SOURCE=2 -fexceptions -fstack-protector --param=ssp-buffer-size=4 -Wformat -Wformat-security -fmessage-length=0 -march=armv7-a -mfloat-abi=hard -mfpu=neon -mthumb -Wno-psabi -fPIC -fvisibility=hidden -fvisibility-inlines-hidden -std=c++0x -Wall -W -D_REENTRANT -fPIE -DVERSION=\"1.0-beta1\" -DROCKPOOL_DATA_PATH=\"/opt/sdk/rockpool/usr/share/rockpool/\"
Andrew Branson
@abranson
Jul 23 2016 22:25
ok so if it build it on linux, i get the proper ROCKPOOL_DATA_PATH. On Windows it's always wrong.
but the url handler problem occurs on both
every time i install the rpm, the handers for text/html, x-maemo-urischeme/http and x-maemo-urischeme/https get switched to the list of every android app.
i'm on 2.0.2.48 though, so it could be a quirk in that. can you check your /usr/share/applications/mimeinfo.cache?