Main Bus File Playback by joelmacx · Pull Request #100 · google/eclipsa-audio-plugin

joelmacx · 2026-02-02T23:27:57Z

Description

Playing audio back through arbitrary system devices causes issues in some DAWs. This work moves file playback to the audio processor chain.

Changes

Add a playback processor for file playback.
Remove dead player and UI code.
Simplify the async work required by the IAMF decoder.

Validation and Acceptance Criteria

Added unit test coverage for the new processor.
DAW testing. Test basic player functionality as well as different state transitions like exporting while playing a file, deleting the processor while buffering, deleting the processor while playing etc.

Validated in:

joelmacx · 2026-02-02T23:38:37Z

common/components/src/SelectionBox.h

    }
    if (selectionBox_.getNumItems() == 1) {
-      selectionBox_.setSelectedId(1);
+      selectionBox_.setSelectedId(1, juce::dontSendNotification);


This was causing hard to trace state changes during UI initialization.

joelmacx · 2026-02-02T23:41:21Z

common/processors/file_output/iamf_export_utils/IAMFFileReader.cpp

+  return seekFrame(frameIdx, abort);
+}
+
+bool IAMFFileReader::seekFrame(const size_t frameIdx,


Updating the seek method to make it cancellable as seeking often requires parsing large amounts of data.

joelmacx · 2026-02-02T23:43:21Z

common/processors/file_output/FileOutputProcessor.cpp

+  FilePlayback fpb = fpbr_.get();
+  fpb.setPlaybackFile("");
+  fpb.setPlaybackCommand(FilePlayback::PlaybackCommand::kPause);
+  fpbr_.update(fpb);


I've learned the hard way that it's much simpler to have processors depend on one repository for their state / commands. This is also a more obvious approach to resetting the player.

joelmacx · 2026-02-02T23:46:53Z

common/processors/file_output/FileOutputProcessor.cpp

  if (FileExport::validateFilePath(
          FileExport::expandTildePath(kIamfPath).toStdString(), false)) {
-    // Clean up the file if it exists
-    std::filesystem::remove(kIamfPath.toStdString());


This line causes difficult problems.

As it relates to the player, when export starts we want to release any handles we have to the file that is to be written. This is done asynchronously (we don't want to freeze the audio thread) as it requires joining multiple threads and doing other cleanup. As it's asynchronous there's no guarantee by the time we reach this line that the file has been relinquished.

What happens if we export to a file that is being held or being played?

I guess it's fine if writing to the file doesn't cause issues, maybe this is handled sensibly by the filesystem somehow. But if it blocks export then we should try and release the handles and then flag for the process to continue somehow.

Handle and filesystem behaviour might differ between windows and macos as well.

We do destroy the reader when export begins here so it should technically be fine to delete this file.

I would prefer to let the host OS handle what to do in case things got out of sync at some point and there is a write to a file currently being read. If this were to occur, my position is that it's better in this case to continue reading from the outdated file than to crash - let the OS handle the file conflicts.

In my testing in MacOS and Windows DAWs there aren't any crashes if you start an export while buffering or while playing.

joelmacx · 2026-02-02T23:54:33Z

rendererplugin/src/screens/ElementRoutingScreen.cpp

  audioElementSpatialLayoutRepository_->registerListener(this);

-  // Tie the horizontal scrollbars together
+  profileSelectionBox_.onChange([this]() {


Moved callback enabling for this component to after all the other initialization is done.

BrandenAvalonCx

I noticed 2 issues when testing:

After seeking to the midpoint of the file, if I change Mix Presentation layout and then press play, the player shows that it will play from the middle of the file, but actually plays from the start again. Probably the file should be played from the middle, where the caret shows.
Seeking in general is a bit tough, since there's an immediate load when you click on the seek bar. Maybe we can start seeking on release / mouse button up? In this way would I be able to see the value I'm scrolling the seek to? Not sure if this worked before, so if not we can make it a future improvement instead.

common/components/src/ExportValidation.h

common/data_structures/src/FilePlayback.h

BrandenAvalonCx · 2026-02-04T21:31:52Z

common/processors/file_output/FileOutputProcessor.cpp

  if (FileExport::validateFilePath(
          FileExport::expandTildePath(kIamfPath).toStdString(), false)) {
-    // Clean up the file if it exists
-    std::filesystem::remove(kIamfPath.toStdString());


What happens if we export to a file that is being held or being played?

I guess it's fine if writing to the file doesn't cause issues, maybe this is handled sensibly by the filesystem somehow. But if it blocks export then we should try and release the handles and then flag for the process to continue somehow.

Handle and filesystem behaviour might differ between windows and macos as well.

joelmacx · 2026-02-11T18:21:22Z

common/processors/tests/IAMFBufferedReader_test.cpp

Superficial test changes as I renamed the type it was testing and adjusted its constructor

joelmacx · 2026-02-11T18:48:53Z

I noticed 2 issues when testing:

After seeking to the midpoint of the file, if I change Mix Presentation layout and then press play, the player shows that it will play from the middle of the file, but actually plays from the start again. Probably the file should be played from the middle, where the caret shows.

Seeking in general is a bit tough, since there's an immediate load when you click on the seek bar. Maybe we can start seeking on release / mouse button up? In this way would I be able to see the value I'm scrolling the seek to? Not sure if this worked before, so if not we can make it a future improvement instead.

Changing the layout properly resets playback position. Seeking is now only submitted after release on drag.

BrandenAvalonCx

I think this looks good, I think the cleanup makes this a bit simpler. Tested with REAPER and playback seems to be working fine. I wouldn't mind a few minor improvements, but I think these can be handled as later work.

I think just a general comment, I wouldn't mind seeing more comments in these larger changes. It may be worth writing a longer comment at the top of the FilePlaybackProcessor explaining the logic and how it works in case we need to make adjustments in the future.

File Playback Processor

dabfae4

joelmacx force-pushed the player-mb-state branch from c920e92 to dabfae4 Compare February 2, 2026 23:33

Cleanup

3ee2285

joelmacx commented Feb 3, 2026

View reviewed changes

joelmacx self-assigned this Feb 3, 2026

joelmacx added 5 commits February 3, 2026 14:08

Missed during cleanup?

f3c1a5d

Rollback some logic removed during cleanup

388c8c1

Register listener by default in ctor

00da01d

String compilation in Windows

9d8329f

String compilation in Windows2

ea82dc4

BrandenAvalonCx reviewed Feb 4, 2026

View reviewed changes

joelmacx added 2 commits February 5, 2026 11:03

Initial refactor task system

77ea699

Develop new worker

3cdcbc5

joelmacx force-pushed the player-mb-state branch from 2519ba9 to 3cdcbc5 Compare February 11, 2026 18:09

joelmacx commented Feb 11, 2026

View reviewed changes

joelmacx added 2 commits February 11, 2026 10:22

Cleanup

20175ec

Clear buffer on empty load

f363769

BrandenAvalonCx approved these changes Feb 11, 2026

View reviewed changes

joelmacx merged commit f58bc28 into main Feb 12, 2026
3 checks passed

joelmacx deleted the player-mb-state branch February 12, 2026 17:18

Conversation

joelmacx commented Feb 2, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Changes

Validation and Acceptance Criteria

Validated in:

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

BrandenAvalonCx left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

joelmacx commented Feb 11, 2026

Uh oh!

BrandenAvalonCx left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

joelmacx commented Feb 2, 2026 •

edited

Loading