Testing redisplay code in batch

[Stefan Monnier <monnier@iro.umontreal.ca>]

I have a test for the Bug#43519 which I'd like to add to our test suite,
but it requires the redisplay to do its job, and in batch mode it seems
that Mr. Redisplay gets quite sloppy.

Does someone have experience with batch-tests of redisplay code?

So far I've only use the "let-bind execute-kbd-macro to t" trick in
order to convince minibuffer functions to actually use a minibuffer
and process events from `unread-command-events`, but here I need to go
further because I the "dummy windows" used in batch mode aren't
good enough.

So I'm ideally looking for a way to convince Emacs's internals to
perform redisplay as if there were "real" windows.  I don't need to
"see/display" the result, OTOH because I only need to test things like
`window-start` and `window-end` after calling `redisplay`.

If that can't be done without major surgery to the C code, has someone
already setup some way to run interactive tests in some virtual
interactive session (I guess using something like a detached `screen` or
`tmux`)?


        Stefan