diff options
| author | Carl Hetherington <cth@carlh.net> | 2018-08-08 23:59:26 +0100 |
|---|---|---|
| committer | Carl Hetherington <cth@carlh.net> | 2018-08-09 22:31:11 +0100 |
| commit | 33c42a2c411d7518d0db5dbe338d0b42192c2730 (patch) | |
| tree | 3096052c309d3f33508abba899fa293768a8e678 /src/lib | |
| parent | f0f4368b7de7e20bed07bc4756c4dd6ba5764808 (diff) | |
Rather dubious fix for crash due to the sequence:
- playlist change
- _have_valid_pieces -> false
- signal butler
- but meanwhile, pass happens which calls setup_pieces and starts emitting
data from time 0
- this new data is not in sync with what's already in the audio ring buffers
Diffstat (limited to 'src/lib')
| -rw-r--r-- | src/lib/player.cc | 8 |
1 files changed, 7 insertions, 1 deletions
diff --git a/src/lib/player.cc b/src/lib/player.cc index c1a6c0f3c..2313e09e5 100644 --- a/src/lib/player.cc +++ b/src/lib/player.cc @@ -594,7 +594,13 @@ Player::pass () boost::mutex::scoped_lock lm (_mutex); if (!_have_valid_pieces) { - setup_pieces (); + /* This should only happen when we are under the control of the butler. In this case, _have_valid_pieces + will be false if something in the Player has changed and we are waiting for the butler to notice + and do a seek back to the place we were at before. During this time we don't want pass() to do anything, + as just after setup_pieces the new decoders will be back to time 0 until the seek has gone through. Just do nothing + here and assume that the seek will be forthcoming. + */ + return false; } if (_playlist->length() == DCPTime()) { |