Open michael-schwarz opened 2 months ago
The description of the escape analysis should definitely at least be amended. I guess it would be possible to have two separate analyses and queries for the two different notions but that might be overkill. In the multi-threaded case they would probably have to amplify each other anyway.
Our notion of escaping would generalize anyway with #1491.
We make inconsistent assumptions about what "has escaped" is supposed to mean:
Escape analysis for thread-local variables ([escape]).
45/05
to fail which requiresMayEscape
to answer true also in single-threaded settings.I came across this when considering #1542, where this distinction leads to a new global variable being created when reading to
g#in
that has not received any values as the value of the variable is still tracked locally.