commit 77bd06ea19b5097a648165b6656457d7b4565a18
author Evan Birenbaum <birenbaum@google.com> Mon Oct 26 20:10:11 2020 +0000
committer CQ Bot <commit-bot@chromium.org> Mon Oct 26 20:10:11 2020 +0000
tree e836fd0bafbcb7bb9da484db35f10a5ddefe17b4
parent d01ad0cd0bd23c1d00a71d12e8e3319447847523

[flatland] Flatland present tokens.

Flatland uses a modified version of "Present2 semantics": instead of
telling clients the total number of presents they have remaining in
various places (the Present() callback, RequestPresentationTimes(),
and OnFramePresented() callbacks all contain this value), Flatland
tells clients they have additional presents through a single
OnPresentTokensReturned() callback.

The general flow is as follows:
- Clients may only call Present() when they have at least one token.
- Calling Present() decrements their token count by one.
- The FrameScheduler will trigger increments of the token count when
  a Present() call is fully process for a Flatland instance.
- Clients will receive additional tokens from the
  OnPresentTokensReturned() event, which is called as soon as a
  present is processed (even if the frame isn't rendered yet).
- Clients may assume they start with one token, allowing a single
  present, primarily for setting up the parent link.

This will have to be reconciled with the "old Present2 semantics".
They return num_presents_allowed in the OnFramePresent() callback,
which we still want to have in Flatland.

Fixed: 57823
TEST=fx test -o flatland_unittests
MULTIPLY: flatland_unittests (fuchsia): 30

Change-Id: I5bde8b292acacb58a90df9fa21a83f38b014a3f3
Reviewed-on: https://fuchsia-review.googlesource.com/c/fuchsia/+/438959
Commit-Queue: Evan Birenbaum <birenbaum@google.com>
Reviewed-by: Adam Gousetis <adamgousetis@google.com>
Reviewed-by: Mikael Pessa <mikaelpessa@google.com>
API-Review: Adam Gousetis <adamgousetis@google.com>
Testability-Review: Mikael Pessa <mikaelpessa@google.com>