V1.9.11

This release includes everything from 1.9.10, which was never published, so
it covers all changes since 1.9.9.

Pro DJ Link

• DJM mixer data now works. Channel faders and VU meters from the mixer
  were not being delivered on some setups. The way STC announces itself on
  the network has been corrected against captures of a working Pioneer
  bridge, and the mixer now sends its data as expected.
• CDJ-2000NXS2 stability. Players could hang, disconnect, or stop
  responding, and STC could crash. Several separate causes were found and
  fixed: STC was sending a keepalive packet to NXS2 hardware that made the
  player open TCP connections to STC at around 25 per second, congesting the
  link; STC was opening a new dbserver connection roughly every 1.7 seconds
  instead of holding one; track analysis was not falling back to NFS when
  dbserver queries came back unusable; and NFS reads were large enough to be
  fragmented across the network. Reported by @trickofdjequip and @Joren2087
  over months of shows, with captures — including side-by-side recordings of
  the official bridge and of working third-party software on the same gear —
  without which none of this would have been findable (#6, #9). Both report
  clear improvements, and both are still seeing issues on this hardware, so
  work continues.
• Bridge identity is now selectable, next to the Pro DJ Link interface
  selector, with an AUTO option that observes the network before
  announcing itself. If a DJM is discovered but its faders and VU meters
  never appear, AUTO is the first thing to try.
• XDJ-XZ support. The XZ presents two decks under a single network
  identity, so only one deck was being tracked, and track data requests were
  being ignored. Both decks now appear. Reported by @elevate5406, whose
  captures made the fix possible (#12).
• The Pro DJ Link view gained a layout cycle (2x2 grid, 4x1 horizontal,
  alternating) and a diagnostics readout showing what STC is doing on the
  network, which makes reporting problems much easier.

LTC / SMPTE

• User bits are now supported on input and output. On output you can set
  a fixed hex value, pass through whatever arrives on the LTC input, insert
  the current date, or send a four-character name. On input the recovered
  value is displayed, including a decimal and date reading when the bits
  contain one. Thanks to @robertmcrostie for the request (#13).
• The output phase is now consistent. The LTC signal previously started
  at an arbitrary point within the frame, different every time the output
  was switched on, which made sub-frame alignment impossible. It is now
  aligned to the timecode clock and compensated for the audio interface's
  reported output latency. Reported by @mungewell (#15, #18).
• Sample rate and buffer size are now shown in the output panel as well
  as the input panel. They were always shared between the two; only the
  controls were missing from the output side.
• Hold on Pause, an option that keeps the LTC carrier running while the
  source is paused, for receivers that drop sync when the signal stops.
  Contributed by @chandz102 (#14).
• Fixed the binary group flags, which could intermittently declare that the
  user bits contained date information when they did not.

MTC

• Fixed a glitch at minute boundaries where the minute could jump for a
  couple of frames on some sources. Reported by @mungewell with enough
  detail to reproduce it exactly (#16).
• Incomplete quarter-frame sequences are now ignored instead of being mixed
  with stale data.
• Latency compensation now follows drop-frame numbering on 29.97 DF sources,
  where it could previously report frame numbers that do not exist.

New inputs and outputs

• LA-Net (LaserAnimation Net-Timecode) input and output. Contributed by
  @i-n-g-o (#11).
• Winamp / WACUP input on Windows, following playback position from a
  running instance, with TrackMap and cue triggers. Also works with players
  that expose the same interface, such as AIMP. Requested by @mixtern, who
  supplied the reference material and tested it across a week of rehearsals
  and a full show (#10).

Other

• TCNet global offset (-2000 to +2000 ms) applied to every layer, to
  compensate for the latency of the receiving system.
• Generator A/B loop — set in and out points on the waveform and loop
  that range, with the audio file looping along with the timecode.
• Track Map "Learn" now works with every source that reports a track,
  not just Pro DJ Link — StageLinQ and Winamp included — and retries when the
  metadata has not arrived yet instead of doing nothing.
• StageLinQ playhead is now derived from the track position rather than
  beat and BPM, so moving the pitch fader no longer shifts it. Contributed
  by @pasque (#17).
• New OSC command for relative jog on the generator, also from @chandz102
  (#14).
• The audio output channel is now remembered between launches instead of
  reverting to 1-2 on startup. Reported by @mixtern (#10).

Thanks

To @i-n-g-o, @chandz102 and @pasque for the pull requests in this release,
and to @robertmcrostie, @mungewell, @elevate5406, @mixtern, @Joren2087 and
@trickofdjequip for the issue reports. Several of these fixes came directly
from recordings, measurements and Wireshark captures that made otherwise
invisible problems obvious — the Pro DJ Link work in particular would not
have been possible without them.