Changes that went uncommitted for too long

Signed-off-by: Andrew Opalach <andrew@akon.city>

1 files changed, 38 insertions, 2 deletions

diff --git a/TODO b/TODO
index 222d6ba..f426d15 100644
--- a/TODO
+++ b/TODO
@@ -41,9 +41,27 @@
  - [x] al_atomic_* -> atomic_*
  - [x] Test pause -> cmv-clear -> add audio only
  - [x] libnaunet zlib wrap
- - [ ] Remove entry->ended from sink.
+ - [x] Runtime and simpler SINK_LOCAL.
+  - [x] Is there any race condition around entry->paused?
+ - [x] Audio buffer local pause.
+ - [ ] fade-out on seek? fade-in on any full cut-out?
+ - [ ] win32 -municode -mwindows build fixes.
+ - win32 resize stuff
+  - https://stackoverflow.com/questions/50898990/reduce-flickering-when-using-setwindowpos-to-change-the-left-edge-of-a-window
+  - https://stackoverflow.com/questions/53000291/how-to-smooth-ugly-jitter-flicker-jumping-when-resizing-windows-especially-drag
+  - https://www.gamedev.net/forums/topic/708865-dxgi-flip-model-flickering-during-live-resize/
+  - https://github.com/Blinue/Magpie/pull/1071
+  - https://gist.github.com/apkipa/20cae438aef2a8633f99e10e0b90b11e
+ - [ ] Why does harfbuz use a different ar than what I define.
+ - [ ] Implement resize to buffer.
+ - [ ] Windows test: Have second client do input simulation while causing Select pause on windows cmd.
+  - [ ] PTS gets cooked at some point (u64)-1.
+ - [ ] Log overall bitrate from FFmpeg.
+ - [ ] Is it possibly to cleanly fold DIRECT_MODE more into nn_multiplex/nn_rpc/nn_packet_stream.
  - [ ] Pass over logging.
+ - [ ] TCP throughput testing.
  - [ ] list_remove().
+ - [ ] Remove entry->ended from sink.
  - [ ] Do ntp like latency evaluation.
  - [ ] packet_cache_v2.
  - [ ] Implement queued.
@@ -51,7 +69,8 @@
  - [ ] Header cleanup.
  - [ ] https://github.com/JuliaStrings/utf8proc/blob/master/utf8proc.h
   - [ ] ustr for ui elements.
-  - [ ] remove wstr.
+  - [ ] Remove wstr.
+   - [ ] Consider usage in win32 sections.
  - [ ] Packet benchmark.
   - Create packet, write bunch of ints, send to another thread, thread adds everything.
  - [ ] Startup timing information.
@@ -105,6 +124,21 @@
  - [ ] Mark entries for removal rather than remove in-place.
  - [ ] More optimal command queue.
 
+// @TODO: Could entry->ended be redundant? What about entry->held?
+//  - The catalyst for this is having to check entry->ended in CLIENT_REMOVE_BUFFERS breaking the ability
+//    to maintain the same state if called consecutively.
+//  - As long as end_entry_and_advance_queue() evaluates target, what is the point of
+//    ended except to say all of an entry's buffers are ended.
+//     (except to guess what the server thinks, which was it's original purpose).
+//  - Knowing if the outputs are paused is clearly necessary but what about it's relationship to held?
+//   - entry->paused basically means preempt any queued skip action.
+//   - On video or audio paused, check if SINK_EMPTY or PAUSED and run target (currently not handled at all).
+//   - Fix SINK_LOCAL relying on entry->paused hint from server (rely on clock state?).
+//    entry->paused reduced to nothing but a hint about if the entry is paused when skipping.
+//    Only consideration is for the sink outputs to never stay paused when playing and vise versa
+//    - Then entry->paused actually doesn't matter (sink state is above entry anyway and we should be able to
+//      rely on the clock per-entry).
+
 === Cache
  - See cache/entry.h.
  - [ ] Don't use file_mapped for large files.
@@ -125,6 +159,8 @@
  - [ ] Internalize "packet cache" logic.
 
 === Codecs
+ - [ ] Cleanup hwaccel stuff.
+  - [ ] Separate ff_decoder from hwaccel specific stuff.
  - [ ] Negotiate codec with sinks.
  - [ ] Better wuffs support.
  - [ ] https://snisurset.net/code/abydos