[jogibear9988]

A try/finally in a later multidimensional array bound can clear earlier bounds from the evaluation stack and produce invalid IL. Record those bounds during collection and spill their evaluated values before constructing the array. The first bound alone does not require this spill.

Multidimensional bounds, ordered side effects, cleanup and failing bounds. Regressions use strict FEC compilation for both Standard and LightExpression.

Validation

Release solution build: zero warnings/errors.

Runtime	TestX groups	Legacy tests
.NET 10	445 passed	1,756 passed
.NET 8	445 passed	1,755 passed
.NET Framework 4.7.2	424 passed	1,639 passed

Measurements

Current master f1d2fb33 → this change, Release, Windows ARM64, .NET 8.0.28 / 10.0.9. Compilation uses prebuilt trees; runtime measures invocation through the same wrapper. Two fresh process rounds with reversed load order, seven samples per round, about 30 ms calibration, one logical CPU and tiered compilation disabled. Tables use the median of the two per-round medians.

The controls below produce correct results on master for their measured inputs. Newly supported trees have no valid before/after execution comparison because master rejects them, produces incorrect results or emits invalid IL. Timings include machine and JIT/code-placement variation; the unchanged add control helps show that variation. These observations do not prove universal performance neutrality.

net8.0 results (master → change)

API / workload	Compile ns	Invoke ns	Compile B/op	Invoke B/op
Standard / add_control	1709.2 → 1724.9	3.1 → 3.2	872.0 → 872.0	0.0 → 0.0
Standard / array_bounds	2527.0 → 2461.4	43.3 → 42.9	960.0 → 960.0	64.0 → 64.0
Standard / array_bounds_first_try	1921.4 → 2102.0	37.5 → 35.9	1560.0 → 1560.0	64.0 → 64.0
Standard / array_bounds_unrelated_try	2539.7 → 2647.7	53.9 → 60.8	1560.0 → 1560.0	64.0 → 64.0
Light / add_control	1773.7 → 1683.4	3.1 → 3.0	872.0 → 872.0	0.0 → 0.0
Light / array_bounds	2114.6 → 2284.6	48.5 → 49.7	960.0 → 960.0	64.0 → 64.0
Light / array_bounds_first_try	2597.4 → 2724.1	54.2 → 48.6	1560.0 → 1560.0	64.0 → 64.0
Light / array_bounds_unrelated_try	2759.2 → 2723.0	49.3 → 57.5	1560.0 → 1560.0	64.0 → 64.0

net10.0 results (master → change)

API / workload	Compile ns	Invoke ns	Compile B/op	Invoke B/op
Standard / add_control	4404.4 → 4269.8	2.5 → 2.6	872.0 → 872.0	0.0 → 0.0
Standard / array_bounds	7004.7 → 6592.7	45.1 → 35.0	960.0 → 960.0	64.0 → 64.0
Standard / array_bounds_first_try	7359.7 → 7866.3	43.0 → 46.7	1560.0 → 1560.0	64.0 → 64.0
Standard / array_bounds_unrelated_try	7199.1 → 7128.9	44.9 → 50.6	1560.0 → 1560.0	64.0 → 64.0
Light / add_control	3735.7 → 4010.5	3.0 → 3.1	872.0 → 872.0	0.0 → 0.0
Light / array_bounds	5162.7 → 5163.2	44.5 → 50.1	960.0 → 960.0	64.0 → 64.0
Light / array_bounds_first_try	6764.0 → 6238.3	43.1 → 47.0	1560.0 → 1560.0	64.0 → 64.0
Light / array_bounds_unrelated_try	5842.7 → 5904.5	42.1 → 41.3	1560.0 → 1560.0	64.0 → 64.0

Emitted IL differs in these measured workloads: none.
No measured allocation increase above 0.1 B/op in these controls.