Fixes from testing sink change, document things

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

1 files changed, 7 insertions, 7 deletions

diff --git a/src/buffer/clock.c b/src/buffer/clock.c
index 508a899..67a229b 100644
--- a/src/buffer/clock.c
+++ b/src/buffer/clock.c
@@ -44,13 +44,13 @@ void camu_clock_seek(struct camu_clock *clock, f64 base, u64 target)
             al_atomic_store(f64)(&clock->tick, tick, AL_ATOMIC_RELAXED);
         }
     } else {
-        // Don't touch clock->pause here for the sake of sync.
-        // Theoretically, the sink could rely on a CLOCK_PAUSED event from an entry
-        // even after it was seeked. That would ultimately maintain sync but the
-        // clock would report the old position in get_pts() before triggering CLOCK_PAUSED,
-        // likely confusing the audio/video buffers of an entry. This is mitigated in
-        // CLIENT_REMOVE_BUFFERS by immediately switching to the target instead of waiting for
-        // a clock_callback().
+        // Don't touch clock->pause here for the sake of sync. This means,
+        // theoretically, the sink could still rely on a CLOCK_PAUSED callback
+        // from an entry even after it was seeked. That would ultimately maintain sync but
+        // the clock would report the old position in get_pts() before triggering CLOCK_PAUSED.
+        // Likely confusing an entry's buffers and causing excessive catchup or delay.
+        // We mitigate this sink-side by immediately switching to a potential target in
+        // CLIENT_REMOVE_BUFFERS, if possible, which is always called before an entry is seeked.
         clock->paused_at = 0.0;
     }
 }