xmn has quit [Read error: Connection reset by peer]
Daanct12 has joined #maemo-leste
ceene has joined #maemo-leste
mardy has joined #maemo-leste
joerg has quit [Ping timeout: 244 seconds]
joerg has joined #maemo-leste
xmn_ has quit [Quit: xmn_]
Twig has joined #maemo-leste
xmn has joined #maemo-leste
vgratian has left #maemo-leste [#maemo-leste]
vgratian has joined #maemo-leste
Pali has joined #maemo-leste
uvos has joined #maemo-leste
<uvos>
Wizzup: no
<uvos>
i ment 5.16 as stated, i was mentioning that i have used this >5.15 version for quite some time and have split the 5.15 patch set in feature branches, not that we should switch to this version in particular
<bencoh>
yeah, 64M, makes the n900 feel bloated :)
<uvos>
its less powerfull than the d4's modem :P
Livio has quit [Ping timeout: 260 seconds]
<Wizzup>
uvos: ok so my plan is the following:
<Wizzup>
1. make a simple py script to quickly restore specific 'groups' of registers
<Wizzup>
2. make some diff between before a call ,during a working call, and after a call (to restore them)
<Wizzup>
3. look into the differences, but only for handset for now
<Wizzup>
some other thoughts I had, you said with the current sound driver design, the kernel turns off entire parts, can we just make some control (on/off) that keeps the whole tree on synthetically, and just set it as part of the ucm?
<Wizzup>
of course it's a hack, but it's way better than any kind of reg writing
<Wizzup>
uvos: like, for this part:
<Wizzup>
>The problem is the that the voice if is considered inactive by the kernel, this is because the kernel is not using this if at all. Because it isent, the modem is using it. this causes all the downstream nodes to be put in off state.
<Wizzup>
or is this what you tried earlier already, with forking the driver?
<uvos>
lol
<Wizzup>
hm?
<uvos>
not you
<Wizzup>
:)
<bencoh>
I'm fiddling with alsa drivers at work as well, and ... alsa can really be a pain sometimes, especially with badly designed drivers :/
<uvos>
i just realized that systemui, to play the alarm ringtone sends a dbus message to hildon-sv-notification-daemon wich then sends a dbus message to system ui to play the sound
<uvos>
"some control (on/off) that keeps the whole tree on synthetically" i tried doing that
<uvos>
but i cant find a way to do so via the api avialable to the codec
<uvos>
(ie the only part we control)
<Wizzup>
hmm
<uvos>
but i might have missed something
<uvos>
in general how this works is pretty cryptic to me, since alsa parses this big struct with string arguments
<uvos>
that arnt documented anywhere
<uvos>
so you have to read quite a bit of code to figure out what anything dose
<Wizzup>
ok, well, let me first fiddle a bit more with these register values and such
<Wizzup>
the hack was just something that occured to me on my walk today
<uvos>
and hildon-sv-notification-daemon is owned by hildon-home
<buZz>
hmmf, seems on d4 nowadays i get -all- sms stuck in tx_queue when i send it, even when i send it directly as a reply
<buZz>
a reboot (and 1 tap on icd2 to get ofono to exist(?)) gets them sent
<Wizzup>
buZz: yes, this is again all and the same ofono problems
<Wizzup>
I plan to look at that /after/ the audio things
<Wizzup>
trust me I run into the same problems every day :)
<buZz>
ah, right, np, it just seems new to me
<Wizzup>
basically almost all ofono issues boil down to ofono not realizing the kernel sent it a message
<Wizzup>
and then it gets it either way later or potentially never
<uvos>
oh god this is terrible
<Wizzup>
this is also why the internet activation sometimes just times out
<Wizzup>
I have made a habit of typically activating the internet, and then 10 seconds later sent myself on my other number a sms
<Wizzup>
that usually makes it activate right away
<Wizzup>
so yeah, this needs fixing
<uvos>
so to activate systemui hildon-sv-notification-daemon dlopens a systemui plugin (yes) and then uses it (so we now have 2 copies of the alarm subsystem of systemui running
<uvos>
and then the code is desinged sutch
<Wizzup>
uvos: freemangordon: might have some ideas/hints on this too, I think he knows how it works quite deeply
<uvos>
that if it it cant decode the sound file (wich is provided by the user mind you) it just errors out
<uvos>
(and passes glib NULL in a bad place too causing a assert)
<uvos>
this is why vibration dosent work
<uvos>
im cerntenly never going to use this alarm system for anything valuble after seeing how it works
<Wizzup>
it works well on the n900 for my for years
<Wizzup>
but yeah, interesting :)
<Wizzup>
so what is our actual problem?
<Wizzup>
the sound file or the glib null ptr?
<uvos>
the decode function fails on any file i give it
<uvos>
not sure why yet
<uvos>
also this yeah
<uvos>
but thats a side effect
<uvos>
of it having poor error checking
<uvos>
the duplicated alarm system in systemui is only used in actdead btw
<uvos>
(same code running twice)
<Wizzup>
is something gstreamer missing?
<buZz>
ah right, i was wondering, the /usr/share/sounds/ looked a bit empty, were the old sounds from nokia not released freely?
<Wizzup>
they are not I think
<uvos>
it dosent use gstreamer
<uvos>
those files would not help
<uvos>
it fails on any file
<uvos>
even those
<Wizzup>
uvos: in any case, great work on digging in so far, sorry to hear the architecture is a bit meh...
<Wizzup>
ok
<buZz>
Wizzup: hmmm, so maybe we need a bunch of new sounds?
<uvos>
anyhow lunch
<uvos>
buZz: yeah
<uvos>
buZz: i suggested using the andoid asop soundset for now
<buZz>
is there a list of sounds we need?
<uvos>
since its apache2
<buZz>
i was considering asking a musician
<uvos>
buZz: sure
<uvos>
idk list
<uvos>
just look on freemantle
<uvos>
*fremantle
<buZz>
hmm, i'll check around a bit
<buZz>
yeah , lemme just install it on my d4 :D apt install fremantle
<buZz>
hehe
<buZz>
hehe well, i can dig up a n900 i guess
<uvos>
or look at the pkgs
<buZz>
not really sure where they went after the last 4 moves
<uvos>
i gues
<buZz>
uvos: or enable inotify logging and run -all- the programs
<buZz>
:D
<buZz>
haha
<buZz>
jk
<buZz>
actually, would be funny
<uvos>
also very slow
<buZz>
some auto-synth that makes a 'new sound' based on the filename of a unfound .wav
<Wizzup>
brb lunch
<buZz>
eet smakelijk
<uvos>
would be a derivatve work
<uvos>
ithen
<uvos>
ie a violation
<uvos>
i would just source some well licenced samples
<buZz>
huh? i dont think so
<uvos>
(ex the android ones)
<uvos>
oh unfound
<uvos>
nvm yeah sure
<buZz>
i mean, a synthesizer can use anything as input, human turning knobs, ascii bytes from filename, etc
<buZz>
yeah
<uvos>
right
<buZz>
buuuut, i think i'll just go make 'some' sounds , just a set of a couple
<buZz>
something with nice long stretched echos or something
<buZz>
as placeholder at least, for stuff i found missing
<buZz>
and just copy those over for as many as i need, so i can build a list
<buZz>
and later ask some musician to make a whole 'theme' just off the filenames
<uvos>
sure yeah, that would be nice
<buZz>
imho leste should have its own sounds
<buZz>
like the basic nokia sounds, triggering nerds worldwide into 'omg you're running leste??'
<buZz>
:D
<uvos>
i do like the ATT model of 500 telephone sphone as has default ringtone
vgratian has left #maemo-leste [Error from remote client]
<uvos>
its not currently the default when you compile sphone on leste, becuse it uses libprofile there
<uvos>
wich has the non existant nokia tune as default ringtone
<uvos>
but if you use sphone on anything else this is default
<buZz>
ah well
<buZz>
maybe i could imitate the nokia sounds on a kazoo?" :D
<buZz>
hahaha
vgratian has joined #maemo-leste
norayr has left #maemo-leste [Error from remote client]
ceene has quit [Ping timeout: 246 seconds]
raub has joined #maemo-leste
norayr has joined #maemo-leste
<norayr>
i didn't add dvorak layout last time, because i don't know how to name the dvorak layout.
<norayr>
in X11 that's again en_US but the variant is dvorak
<norayr>
and we in hildon input method plugins have
<norayr>
the xml file named en_US
<norayr>
should i do en_DV ? but that would mean wrong country.
<uvos>
so decoding works by passing a filenname to a com.nokia.NsvDecoder and then making assumptions that it saves a equivalent wav file in a magic folder (it dosent)
<uvos>
also seams realy racey
<Wizzup>
uvos: hm... so the wav is to not have to decode in the future?
<Wizzup>
is that the thought?
xmn has joined #maemo-leste
<uvos>
yeah but no idea why they do this in a seperate process via dbus
<uvos>
just call gstreamer youself
<uvos>
they also dbus_g_proxy_begin_call
<uvos>
and then assume the file exists right after
<uvos>
i dont see how thats not really racey
<Wizzup>
hmmmmm
<Wizzup>
which code is this?
<uvos>
unless im missing somehting
<uvos>
nsv_decoder_decode in hildon-plugins-notify-sv
<Wizzup>
it looks like _nsv_decoder_decode_cb is called once the call is gone, which then call the real callback?
<uvos>
sure
<Wizzup>
even though _nsv_decoder_decode_finished_cb seems empty
<uvos>
expect thats not how it works
<Wizzup>
but I guess maybe that gets override somewhere
<uvos>
so the code that plays the sound assumes the wav exists allready
<uvos>
and fails hard when it dosent
<uvos>
(ie using NULL)
<uvos>
and the call to decode is issued by the ui
<Wizzup>
aha
<Wizzup>
maybe the idea is to isolate errors in the decode sw so another process somehow
<Wizzup>
just thinking out loud
<uvos>
yeah
<Wizzup>
that is not bad per se
<uvos>
i dont know what process implements com.nokia.NsvDecoder atm
<Wizzup>
want me to grep?
<uvos>
Error: GDBus.Error:org.freedesktop.DBus.Error.NameHasNoOwner: Could not get PID of name 'com.nokia.NsvDecoder': no such name
<Wizzup>
(nsv-decoder-service:9129): GStreamer-WARNING **: 16:28:42.426: 0.10-style raw audio caps are being created. Should be audio/x-raw,format=(string).. now.
<Wizzup>
(nsv-decoder-service:9129): GLib-GObject-WARNING **: 16:28:42.428: ../../../gobject/gsignal.c:2524: signal 'new-decoded-pad' is invalid for instance '0x669170' of type 'GstDecodeBin'
<Wizzup>
so if you run the program manually (from libexec) and then while it runs, run the mdbus2 stuff, you should see something similar
<Wizzup>
"decodebin2" has been renamed to "decodebin", with similar API. Note that there is no longer a "new-decoded-pad" signal, just use GstElement's "pad-added" signal instead (but don't forget to remove the 'gboolean last' argument from your old signal callback functino signature).
<Wizzup>
do you want me to try to fix this on my d4?
<Wizzup>
or are you on it?
<uvos>
im afk ffor a couple of hours
<uvos>
ill do it after
<Wizzup>
I will look at it now
<uvos>
or you can if you like
<Wizzup>
(nsv-decoder-service:13166): GStreamer-WARNING **: 16:41:48.041: 0.10-style raw audio caps are being created. Should be audio/x-raw,format=(string).. now.
<Wizzup>
0:00:01.256317139 13166 0x63ffa0 WARN basesrc gstbasesrc.c:3583:gst_base_src_start_complete:<filesrc0> pad not activated yet
<Wizzup>
or figure out an entire different way to do all of the routing that exists
<Wizzup>
in any case we're not there yet
<freemangordon>
Wizzup: ah, so I didn;t get what uvos meant it seems
<freemangordon>
ok
<freemangordon>
do you want me to do a changelog/release for hildon-plugins-notify-sv?
<sicelo>
Hehe, @alarm problems on n900. That alarm is the ONLY alarm which can be relied on to wake one up even when device is off, in the entire mobile linux space (including Android, at least I've used about 6 different android phones/versions over the years, and they never could wake up for an alarm, unless I just had bad luck)
<Wizzup>
freemangordon: yes, I think so
<freemangordon>
ok, will do
<freemangordon>
sicelo: I was away, what problems to you mean?
<freemangordon>
*do you
<Wizzup>
it was regarding comments from uvos earlier I think
<Wizzup>
+ the gst stuff I fixed
<uvos>
the implementaion of alarms is not so great
<uvos>
its fairly fragile
<uvos>
and pulls in a lot of stuff for farily limited reasons
joerg has quit [Excess Flood]
<freemangordon>
maybe I shall read the backlog about that
<sicelo>
freemangordon - I was reading backlog, yes.
_j has joined #maemo-leste
_j is now known as joerg
<freemangordon>
uvos: unless you can summarize what you think is bad
<freemangordon>
otherwise I igree there are issues
<freemangordon>
*agree
<freemangordon>
one of them is that it is possible that you cannot stop an alarm immediately after you start it
<freemangordon>
alarm tone that is
<Wizzup>
uvos: btw
<freemangordon>
so deffinitely there is a field for improvement
<Wizzup>
these are the exact bit-by-bit differences
<Wizzup>
I think not all the bits are necessary for the calls
<Wizzup>
the mic gain ones probably don't matter
<Wizzup>
(as they are just a bit diff)
<sicelo>
You're comparing on/with Android?
<Wizzup>
CPCAP_REG_VAUDIOC + CPCAP_REG_CC are probably the 'kernel route' related ones
<Wizzup>
maybe also the SDAC and SDACDI ones
<freemangordon>
uvos: but, besides that, alarm on n900 is what I am using to wake me up for the last 11 years. It has *NEVER* failed, no matter what
<bencoh>
:)
<freemangordon>
well, not after I pushed alarms queue fix in CSSU :)
l_bratch has quit [Quit: Leaving]
l_bratch has joined #maemo-leste
<norayr>
freemangordon, may it be you know what to do if i'd like to create a dvorak layout?
<norayr>
i didn't add dvorak layout last time, because i don't know how to name the dvorak layout.
<norayr>
in X11 that's again en_US but the variant is dvorak
<norayr>
and we in hildon input method plugins have the xml file named en_USthe xml file named en_US.
<norayr>
should i do en_DV ? but that would mean wrong country.
<norayr>
(for some reason pasted twice)
<norayr>
i mean my problem is the naming of the file.
<norayr>
i don't see anything like 'variant' in our xml files.
<Wizzup>
I don't think this is how it works
<Wizzup>
isn't there a layout for en_US ?
Danct12 has quit [Quit: Quitting]
Danct12 has joined #maemo-leste
<uvos>
Wizzup: him dosent do variants afaik
<uvos>
freemangordon: it working on fremantle is a poor indicator of anything, since it was dead on arrival and never chainged.
<uvos>
freemangordon: i just noticed that the alarm system needs a large amount of stuff to work via lots of dbus round trips and is not fault tollerant at all, because it either dose no error checking and ends up asserting in glib-critical because its using NULL pointers or it just bails
<uvos>
ie: sound files missing? assert, decode /gst not working on user file? bail, new pulse audio so one of your modules dident load? just bail. systemui dosent respond in timeout? bail
<uvos>
etc
<norayr>
Wizzup, yes there is a layout for en_US
<norayr>
but that's a qwerty layout
<norayr>
it's really hard to type qwerty, especially on big pinephone on-screen kbd.
<norayr>
so i need my native dvorak layout.
<norayr>
i am not sure how to name the xml file.
<uvos>
id really expect a alarm system to at least try and continue if it cant find a sound file or cant set the volume etc
<norayr>
we have en_US
<norayr>
and en_GB
<norayr>
in X11, however that's
<norayr>
file named us, and different layouts in it, for english in usa.
<norayr>
one of those is qwerty, one is dvorak.
<norayr>
(on a computer i use en_US(dvp) which means en_US, but the layout is programmer's dvorak.
Danct12 has quit [Quit: Quitting]
<norayr>
let's say in Bulgarian you have several variants:
<norayr>
bds, phonetic, bas_phonetic, bekl, latin
<norayr>
ok i guess i'll just name the file somehow.
<norayr>
en_dvorak, and let's see how it'll compile.
<Wizzup>
uvos: ok, with some hacks I reliably have call audio now upon calls
Danct12 has joined #maemo-leste
<Wizzup>
but it's revealing some other problems, but that's all fine, teporary
<Wizzup>
norayr: well, yes, variant is what you need to add, no?
<uvos>
Wizzup: the problem is not his x layout
<uvos>
its the him vkb layout
Danct12 has quit [Client Quit]
<Wizzup>
how does it work on fremantle?
<uvos>
no idea
<uvos>
dose it work?
<Wizzup>
I don't know, I was never masochistic enough to try :P
<Wizzup>
so one of the funny things about sphone module order is that it will either do pulse first and then my hack, or reverse
<Wizzup>
but ideally it would be pulse first, then the hack, but when the call ends, reverse order
<Wizzup>
uvos: it looks like our usecase with the dsp
Livio has joined #maemo-leste
<Wizzup>
freemangordon: did you rebuild the notify sv thing?
<Wizzup>
uvos: not sure if it will actually work for us, but it looks interesting
<Wizzup>
In order for this aforementioned route to get triggered, DAPM needs to find a valid endpoint which could be either a sink or source widget corresponding to playback and capture path respectively.
<uvos>
i mean now clock-ui works
<freemangordon>
Wizzup: no, sorry, will do in tomorrow
<uvos>
but it hangs until the audio file stops playing
<uvos>
obvious side effect of calling the main loop
<uvos>
...
Livio has quit [Ping timeout: 258 seconds]
<Wizzup>
uvos: ok, well, it's good for testing alarms, but we'll have to fix that soon :)
<Wizzup>
uvos: alt. can we make some fake Mux perhaps
<uvos>
uvos: Project ERROR: Unknown module(s) in QT: multimedia
<uvos>
ah you did
<uvos>
and it failed
<Wizzup>
ah
<Wizzup>
so we lack a pkg :)
<uvos>
yeah minor porblem
<Wizzup>
lmk what to add to control and I'll do it
<Wizzup>
no need ot pr
<uvos>
libqt5multimedia5 to build-depends i gues
<uvos>
its a bit wierd libqt5multimedia5 dosent have a devel equivalent
<Wizzup>
qtmultimedia5-dev
<uvos>
right
<uvos>
thats probubly it
<uvos>
nice package nameing there
<Wizzup>
\o/
<Wizzup>
uvos: really nice that the alarm works
<uvos>
yeah
<Wizzup>
/usr/include/glib-2.0/glib/gtypes.h:32:10: fatal error: glibconfig.h: No such file or directory #include <glibconfig.h>
<uvos>
ah yeah
<Wizzup>
wait, did you remove glib?
<uvos>
it still uses glib but that was just included by depens via gst
<uvos>
no
<Wizzup>
ah
<Wizzup>
so do we need some glib dev?
<Wizzup>
pkg?
<Wizzup>
libglib2.0-dev ?
<uvos>
should be yes
<Wizzup>
uvos: btw I think there was a suggeston in the dapm somewhere that there could be a second almost dummy codec for this dai to dai without cpu thing
<Wizzup>
uvos: still a problem
<Wizzup>
uvos: I think some QT += stuff is required
<Wizzup>
I guess PKGCONFIG
<uvos>
hmm
<uvos>
but it builds locally
<uvos>
i dont se how that would change it
<Wizzup>
PKGCONFIG += glib-2.0
<uvos>
(building remote)
<Wizzup>
let me just try this
<uvos>
a yes
<uvos>
clean building it local dosent work anymore
<Wizzup>
figured
<Wizzup>
that worked
mardy has quit [Quit: WeeChat 2.8]
<uvos>
great
<Wizzup>
what alarm sound do you use?
<Wizzup>
are the defaults installed?
<uvos>
yes
<Wizzup>
ok, let me try it
<uvos>
should play
<uvos>
when you select it
<kona>
bedaboo-ba-bedaboo-ba-bedaboo-ba-be?
<uvos>
no
<kona>
not a terribly good alarm sound anyway :)
<Wizzup>
uvos: oh, didn't try, I usually have the device on silent :D
Livio has quit [Ping timeout: 258 seconds]
<Wizzup>
uvos: it might not even be that hard to switch away from graph card I think?
uvos has quit [Ping timeout: 246 seconds]
norayr has quit [Ping timeout: 252 seconds]
vgratian has quit [Ping timeout: 258 seconds]
norayr has joined #maemo-leste
<Wizzup>
uvos: I wonder if we can't just set something like 'ignore suspend' on the properties we want to keep on (unless overriden by say hi fi)
<Wizzup>
so our graph-card 'usage' is just in the dts right, and that just sets things up somehow?
Pali has quit [Ping timeout: 246 seconds]
Oksanaa has joined #maemo-leste
<kona>
excuse my ignorance, but what is the difference between the leste-image-pinephone and the leste-image-pinephone-dev?
<Wizzup>
the latter has the -devel repos enabled
<kona>
cool.
<kona>
i'm about to try something wacky
<kona>
i left my (only?) sd-card adapter 90 minutes away, so i'm going to try writing the latest maemo image to sd from mobian on emmc.