Extract TrackCollectionManager from Library

[uklotzde]

As discussed on Zulip

• Extract TrackCollectionManager from Library.
• Route all modifying requests of Tracks through TrackCollectionManager to keep both the internal TrackCollection and all external track collections synchronized.
• Fix shutdown order and process to avoid spurious crashes in GlobalTrackCache when exiting the application, caused by delayed destruction of internal TrackPointer instances in various components.
• Only formal, no functional changes besides bug fixing!

Marks as [WiP] to allow rebasing on master. Otherwise merge conflicts may arise and complicate the review of the actual changes.

[daschuer]

Marks as [WiP] to allow rebasing on master. Otherwise merge conflicts may arise and complicate the review of the actual changes

So this should not be reviewed yet?
In case of conflicts you can just merge master and resolve conflicts. GitHub shows than only the original changes and this can still be reviewed.

[Holzhaus]

Thanks for trying to clean up the database access.

This compiles and all tests still pass. However, I get a segfault when starting Mixxx:

[...]
ALSA lib pcm_dmix.c:1108:(snd_pcm_dmix_open) unable to open slave
ALSA lib pcm_dmix.c:1108:(snd_pcm_dmix_open) unable to open slave
connect(2) call to /dev/shm/jack-1000/default/jack_0 failed (err=No such file or directory)
attempt to connect to server failed
Warning [Main]: ControlDoublePrivate::getControl returning NULL for ( "[AutoDJ]" , "enabled" )
Segmentation Fault (core dumped)

$ coredumpctl -1 info
           PID: 153595 (mixxx)
           UID: 1000 (jan)
           GID: 1000 (jan)
        Signal: 11 (SEGV)
     Timestamp: Thu 2019-11-21 13:57:32 CET (52s ago)
  Command Line: ./mixxx
    Executable: /data/jan/Projects/mixxx/cmake_build/mixxx
 Control Group: /user.slice/user-1000.slice/session-1.scope
          Unit: session-1.scope
         Slice: user-1000.slice
       Session: 1
     Owner UID: 1000 (jan)
       Boot ID: 7db033707c384c9ca8d05dc25469917d
      Hostname: jan-desktop
       Storage: /var/lib/systemd/coredump/core.mixxx.1000.7db033707c384c9ca8d05dc25469917d.153595.1574341052000000000000.lz4 (truncated)
       Message: Process 153595 (mixxx) of user 1000 dumped core.

                Stack trace of thread 153595:
                #0  0x00007ff566ef8534 n/a (n/a)

[uklotzde]

69faf56 fixes a double free issue that was caused by introducing parented_ptr in MixxxMainWindows. Maybe this could be the cause for the deadlock?

operator T*()in parented_ptr is convenient but dangerous, because it allows to invoke delete on wrapped pointers without any warning!

[daschuer]

Yes, maybe there is s way to delete the delete operator.

void operator delete (parented_ptr<T> ptr) = delete

Or delete the void* cast

operator void*() = delete

[uklotzde]

Overloading operator delete does not work.

[daschuer]

Why not in our case? I have just read that it should work.

Did you try as member function

void operator delete(void* ptr) = delete;

Or keep it and do nothing in the body.

[uklotzde]

Ok, then this might be a special case for QMainWindow. The debug assertion in the destructor of parented_ptr was triggered.

But, anyway, deleting a parented_ptr is unintended. Pointers should be managed and owned only by a single authority.

[uklotzde]

Now you have to explicitly obtain the plain pointer from parented_ptr if your want to delete it manually.

[uklotzde]

To use the function pointer variant of QMetaObject::invokeMethod() was not really helpful. I didn't check myself that this has been added "recently" in Qt 5.10 and is not available in earlier versions.

[uklotzde]

Build errors are fixed.

I'm not able to reproduce any deadlock on shutdown.

[daschuer]

I will watch this a bit and if this does not appear again, we can merge.

[Be-ing]

Ping @daschuer, ready for merge? Let's not keep this open indefinitely.

[daschuer]

I did some more test without any deadlock. So we can merge it now.
Thank you @uklotzde

[uklotzde]

Thank you all for your efforts and trust. I'll try to keep the following PRs smaller for easier review.