WIP: Continue work on input_cache #1047
Conversation
| @@ -114,7 +114,7 @@ struct playlist { | |||
| int GetCurrentPosition() const noexcept; | |||
|
|
|||
| gcc_pure | |||
| int GetNextPosition() const noexcept; | |||
| int GetNextPosition(unsigned index) const noexcept; | |||
There was a problem hiding this comment.
Please document this - the method is currently completely undocumented, that's bad already (sorry...)...
| else if (queue.IsValidOrder(current + 1)) | ||
| return queue.OrderToPosition(current + 1); | ||
| else if (queue.IsValidOrder(current + index)) | ||
| return queue.OrderToPosition(current + index); | ||
| else if (queue.repeat) | ||
| return queue.OrderToPosition(0); |
There was a problem hiding this comment.
Incomplete: in repeat mode, this should properly calculate the overflow
| @@ -114,7 +114,7 @@ struct playlist { | |||
| int GetCurrentPosition() const noexcept; | |||
|
|
|||
| gcc_pure | |||
| int GetNextPosition() const noexcept; | |||
| int GetNextPosition(unsigned index) const noexcept; | |||
There was a problem hiding this comment.
Can we have index=1 as default value, so existing callers don't need to be migrated?
Maybe rename the parameter to offset?
| @@ -80,7 +80,7 @@ PrefetchSong(InputCacheManager &cache, const char *uri) noexcept | |||
| static void | |||
| PrefetchSong(InputCacheManager &cache, const DetachedSong &song) noexcept | |||
| { | |||
| PrefetchSong(cache, song.GetURI()); | |||
| PrefetchSong(cache, song.GetRealURI()); | |||
There was a problem hiding this comment.
The commit message isn't good - it doesn't explain what needs to be fixed, and why this change fixes the issue.
| @@ -225,6 +228,7 @@ Partition::OnIdleMonitor(unsigned mask) noexcept | |||
| void | |||
| Partition::OnGlobalEvent(unsigned mask) noexcept | |||
| { | |||
| // TODO This doesn't seam to trigger when the random property of the queue is changed, this is needed for the pre caching | |||
There was a problem hiding this comment.
A "sync with player" is only needed if the one next song changes; the player doesn't care for anything but the current and the next song. And the next song only if it has already finished decoding the current one; then it will start decoding the next one. That's why.
(This pr is mostly for seeing what this fork is doing since that is hard to see on github.)
My fixes are mostly wip, see the TODOs for further information.
In my tests I was able to use an unstable sshfs connection (make shure to use the timeout options so that the fs calls actually fail).
I also tested with curl / webdav (Remove /src/input/cache/Manager.cxx:97-98 for that), but that somehow still fails even if cached, didn't look further into it.
I will continue working on this, but probably not fast. So help with the TODOs (also ideas how to fix them) and more are welcome.