This is comparable with the latency of Radek's GSM voice routing program, and I'm hoping that increasing it will allow pulseaudio to use less CPU, and maybe avoid some of the current ghosting artefacts.
It's a C version of Ruby code (show_msm_messages.rb) from the qcombbdbg project, as brought to the GTA04 mailing list's attention by Denis 'GNUtoo' Carikli in http://lists.goldelico.com/pipermail/gta04-owner/2012-November/003269.html.
Following Radek's identification of the problem being how sqlite handles ~status, where status is a quint64, this change masks off the bits of ~status that have no meaning, leaving a much smaller number for sqlite to deal with. For me this really appears to solve the Draft Message problem.
…modem', 'nj-codebase-pruning' and 'nj-audio-oldrouting' into nj
Specifically these: Nov 20 22:09:04 neo Qtopia: QString::arg: Argument missing: "1 missed", @/Communications/Calls/MissedCalls Nov 20 22:09:04 neo Qtopia: QString::arg: Argument missing: "1 new", @/Communications/Messages/NewMessages These arise because the "faen"-derived themes have special cases for 1 missed call and 1 new message - presumably for translation into languages where the 1 case is different from N != 1. All those places have an unnecessary <trarg>, which causes the logs, and which this commit removes.
…eartbeat Sadly, and ironically, this meant that ticking the "Push Enabled" checkbox for an IMAP account could actually do _more_ data transfer than without "Push Enabled". For example I used to have a checking interval of 10 minutes, but enabling "Push Enabled" reduced that to 2 minutes, because my server emits an "* OK Still here" heartbeat every 2 minutes on the IDLE connection.
For example, signed email from KMail, that then goes through a mailing list, ends up with this structure: multipart/mixed multipart/signed multipart/alternative text/plain text/html application/pgp-signature text/plain (mailing list trailer)
My ALSA state rework accidentally lost recording function, e.g. as used by the Voice Notes application. This commit recreates it. I thought it might be possible for the default Media state - normally MediaSpeaker - to handle both playback and recording. But this produces a nasty feedback loop, because Pulseaudio is by default always looping the default source (i.e. the microphone) back to the default sink (i.e. the speaker). Therefore the default MediaSpeaker can't have the microphone switch open, and therefore we need a separate state, MediaRecording, to handle recording. How then, does an application such as Voice Notes get the MediaRecording state installed? My guess is that it happens automatically when any applications uses a QAudioInput, so long as all the states have their Input/Output capabilities correctly defined.
… Device or resource busy" output from alsactl Previously we'd kill all soundcard users if there was any output at all from alsactl. Killing soundcard users is possibly a bad idea in any case, but it's especially so when the output is actually benign. For example, it means that the Media Player dies when headphones are inserted or removed. Regarding the "Codec Operation Mode" setting, Neil Brown recently wrote: > The twl4030 audio codec has 2 operation modes. > In one the PCM (digital) connections to the bluetooth and Modem are > disabled. In the other they are enabled. > The first is sometimes described as "multimedia player" mode while the > seconds is "voice" mode. I think you a limited to lower sample rates in > voice mode. > > You cannot switch modes while the device is active. i.e. nothing can have > the device open to play or record. > > This error just tells you that it cannot switch mode. It shouldn't stop > other settings from being applied. > > On the A3 you should never need to switch modes (unless you are playing with > using the PCM path to get sound to/from the bluetooth). > On the A4 you need to either leave it in the second mode (which probably > wastes power, and probably forces low data rates) or switch on the second mode > at some safe point - maybe between stopping the ring and answering the call, > or maybe before making the phone ring. On the A3 board, all the possible ALSA states (i.e. all except *hw.state) have 'Codec Operation Mode' = 'Option 1 (audio)', so there are no transitions that actually switch the mode, and it is not a problem for that setting to fail. Therefore we should allow that output and _not_ go killing soundcard users. This change allows media playback to continue when headphones are inserted or removed. It also seems to make Voice Notes work more reliably, and I suspect it might help with call audio too. (Unfortunately this change does nothing yet for the A4 board. On the A4 things are more complex, as Neil Brown says above, because there there are some ALSA state transitions that _do_ need to change the codec operation mode. So for A4 more thought and work are needed.)