Could LambdaExpression.analyseExceptions() cause problematic side-effects?

[stephan-herrmann]

Method LambdaExpression.analyseExceptions() runs flow analysis ahead of time.

As per #1181 (comment) I wonder, whether this could cause side-effects outside the lambda itself (the lambda is a throw-away AST)?

What are the side effects caused by flow analysis:

• local FlowInfo, FlowContext not a problem 👍
• local bits in ASTNode.bits, not a problem, as that's only the AST copy 👍
• synthetics in outer scopes / types, for outer emulation, synthetic arguments etc.

Can it happen that along the lines of (3) analyseExceptions() will leave side effects outside the copy, which could have an ill effect on the final flow analysis?

[stephan-herrmann]

Could eclipse-platform/eclipse.platform.releng.aggregator#1940 actually be a witness of non-local side-effects caused by analyseExceptions()?

Observed changes:

1. Generated attribute got reordered
2. There is a reorder in lambda parameters.

What determines such ordering? Aren't variable indices managed by enclosing scopes? Could #1181 have changed the order in which such things are registered?

If there is such connection, it wouldn't indicate a bug (yet), but might help us understand the actual impact of things happening on lambda copies, effects that should stay in Vegas, but something might still leak.