bug: loss of input on ReleaseTerminal / Bubbletea shutdown

[knz]

It is not currently possible to integrate Bubbletea-based programs in scripts or unit tests that buffer terminal input across multiple runs of the Bubbletea event loop.

The specifics are explained here: https://dr-knz.net/bubbletea-control-inversion.html

In summary, the input reader function is "greedy" but drops/forgets input event messages during ReleaseTerminal(). This is where the input is lost.

The text linked above outlines a solution:

• [x] invert control of the input reader function, i.e. this PR: https://github.com/charmbracelet/bubbletea/pull/569
• [ ] synchronise the reading from the external input with Update calls.
• [ ] for the benefit of programs (like Bubbline) that need to share control of the terminal, introduce a new API to suspend the Bubbletea event loop.

[knz]

cc @muesli @meowgorithm

[chabad360]

Is there any update on this? This seems to happen pretty consistently when mouse input gets involved.

[knz]

@chabad360 can you say more? What are the symptoms?

[chabad360]

After clicking around for a bit (an easy way to speed this up is to click outside the terminal), the entire system stops receiving input but does still respond to other things like resizing the window. If I force kill the program with pkill, the terminal no longer accepts input.

Now that I'm rereading the issue, it seems different, but still related.

[knz]

Thanks. I think you are experiencing the symptoms of a different problem, but i believe it could also improve with this change https://github.com/charmbracelet/bubbletea/pull/569

[knz]

And also this change https://github.com/charmbracelet/bubbletea/pull/570

[chabad360]

hmm, go mod replaceing that branch in doesn't seem to help...

[malko]

Is there any plan to address this issue in a near future ? Side note great and interesting explanation of @knz in the provided link.