markdown · 1243 bytes Raw Blame History

Noted Items

Deferred items that came up while finishing Sprint 29.10 cleanup work and starting the full Sprint 29 audit:

  • Audit and harden descriptor-backed integer(16) formatted section reads at the backend/harness boundary. The lowering gap is now closed for allocatable section destinations with real runtime bounds/strides, and dedicated fixture-backed audits cover IR plus O1+/high-opt runtime behavior, but O0 still exposes the existing large-frame slot-addressing backend issue in some cases and the external whole-array fixture still returns exit -1 under capture_from_path(...Stage::Run...) even though direct CLI O1/O2/O3 runs succeed.
  • Parser gap: typed character array constructors using an explicit type-spec inside brackets (for example [character(len=20) :: "a", "b"]) still fail to parse in at least one real-world-style source shape. The full-sprint audit tripped this while building the fpm-inspired realworld_suffix_scan.f90 reproducer, which is currently written in a more conservative source form instead of the typed constructor spelling.
  • Revisit ambitious array/vectorization-style integer(16) rewrites only after the scalar/runtime ABI surface is fully closed and audited.
View source
1 # Noted Items
2
3 Deferred items that came up while finishing Sprint 29.10 cleanup work and
4 starting the full Sprint 29 audit:
5
6 - Audit and harden descriptor-backed `integer(16)` formatted section reads at the
7 backend/harness boundary.
8 The lowering gap is now closed for allocatable section destinations with real
9 runtime bounds/strides, and dedicated fixture-backed audits cover IR plus
10 O1+/high-opt runtime behavior, but O0 still exposes the existing large-frame
11 slot-addressing backend issue in some cases and the external whole-array
12 fixture still returns `exit -1` under `capture_from_path(...Stage::Run...)`
13 even though direct CLI O1/O2/O3 runs succeed.
14 - Parser gap: typed character array constructors using an explicit type-spec
15 inside brackets (for example `[character(len=20) :: "a", "b"]`) still fail
16 to parse in at least one real-world-style source shape.
17 The full-sprint audit tripped this while building the fpm-inspired
18 `realworld_suffix_scan.f90` reproducer, which is currently written in a more
19 conservative source form instead of the typed constructor spelling.
20 - Revisit ambitious array/vectorization-style `integer(16)` rewrites only after the
21 scalar/runtime ABI surface is fully closed and audited.