[daveedvdv-nvidia]

Fixes #521.

When extract_value_from_constant copies a string literal (a ck_string
constant) into the storage being initialized, it maps that storage back to the
constant (map_stack_bytes). Callers that copy into aggregate elements undo
that mapping afterward, but process_dyn_init_work did not do so for the
dik_constant initializer of a local array. So when the same storage was
initialized again (e.g., in the next iteration of a loop, as in the issue's
example), the interpreter tried to map an existing key and asserted
"duplicate map key in interpreter".

process_dyn_init_work now unmaps the array storage after a ck_string
initializer has been copied. It does so only if the storage is actually mapped
to that constant, because extract_value_from_constant can return before
installing the mapping.

The new test regressions/gh-521.sft.cpp covers the issue's example, an array
modified on each iteration, and const char[], char16_t, and wchar_t
arrays, in C++14 and C++20 modes.

Summary by CodeRabbit

• Bug Fixes
  • Fixed an internal error that could occur when local character arrays initialized from string literals were initialized again in a loop.
  • Repeated initialization now works as expected for mutable and const arrays, including narrow and wide character types, in C++14 and C++20.

[coderabbitai]

No actionable comments were generated in the recent review.

Walkthrough

The interpreter now removes a destination-storage mapping when it still points to an extracted string constant. Regression tests cover repeated local array initialization from string literals during constant evaluation in C++14 and C++20 modes.

Changes: Repeated string initialization

Layer / File(s)	Summary
Mapping cleanup and regression coverage src/interpret.c, tests/tests/regressions/gh-521.sft.cpp, tests/tests/regressions/.gh-521.rto/*, src/pending_changes/gh-521-string-init-local-array-loop.txt	After aggregate extraction, the interpreter removes the destination-storage mapping only when it still points to the extracted string constant. Regression checks cover repeated narrow, UTF-16, and wide array initialization, including per-iteration mutation. The test fixtures run in C++14 and C++20 modes.

Estimated code review effort: 2 (Simple) | ~10 minutes

Change: Bug fix · Severity of issue fixed: Low

Suggested reviewers: wchilders-nvidia, daveedvdv

Merge Risk: Minimal · up to f0c2e

The change fixes an internal compiler error during repeated constexpr initialization of local arrays from string literals and adds regression tests. No merge-blocking risk is identified.

Pre-merge checks: 4

Check name	Status	Explanation
Description Check		Check skipped - CodeRabbit’s high-level summary is enabled.
Title check		The title names the main change: repeated constexpr initialization of a local array from a string literal, which matches the regression fix for GH #521 in the interpreter's dynamic initializer handlin…
Linked Issues check		Issue #521 requires that a constexpr loop which initializes a local char array from a string literal more than once must not trigger the "duplicate map key in interpreter" assertion. In src/interpret.…
Out of Scope Changes check		All whole-PR changes relate to issue #521. The src/interpret.c change is the direct fix for the duplicate-map-key assertion. The regression test gh-521.sft.cpp and its .rto fixture files (default.1.1.…

Full details: Title check

Explanation

The title names the main change: repeated constexpr initialization of a local array from a string literal, which matches the regression fix for GH #521 in the interpreter's dynamic initializer handling and the new regression test. The title is specific and accurate, and it does not describe anything absent from the changeset.

Full details: Linked Issues check

Explanation

Issue #521 requires that a constexpr loop which initializes a local char array from a string literal more than once must not trigger the "duplicate map key in interpreter" assertion. In src/interpret.c, process_dyn_init_work (dik_constant and dik_nonconstant_aggregate branch) now calls get_stack_bytes on the destination storage after extract_value_from_constant. If the storage maps to the same ck_string constant, the code calls unmap_stack_bytes. This removes the stale reverse mapping before the next initialization of the same storage. The guard on the mapping covers the case where extract_value_from_constant returns before installing a mapping. The new test tests/tests/regressions/gh-521.sft.cpp contains the issue's reproducer shape (f(), char a[2] = "a" in a loop) with a static_assert, and it runs in --c++14 and --c++20 modes. It also covers a modified array across iterations (g()) and const char[], char16_t, and wchar_t arrays (h()). The expected values in the static_asserts were checked by hand and match the loop semantics. No coding requirement remains unmet. Automated test execution results were not available to this review, so the runtime pass of the test is not independently confirmed.

Full details: Out of Scope Changes check

Explanation

All whole-PR changes relate to issue #521. The src/interpret.c change is the direct fix for the duplicate-map-key assertion. The regression test gh-521.sft.cpp and its .rto fixture files (default.1.1.txt, default.2.1.txt) verify that fix in C++14 and C++20 modes. The pending_changes entry documents the fix. No unrelated file or hunk was identified in the reviewed diff.

Finishing Touches

Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

• X
• Mastodon
• Reddit
• LinkedIn

Recent review info

Run configuration

• Source mode: Auto
• Repository: edgcpp/compiler
• Branch: main
• File: .coderabbit.yaml
• Commit: 8fa21be
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Advanced
• Run ID: 202e3c04-2515-4378-8cd7-725df483da30

Commits

Reviewing files that changed from the base of the PR and between 8fa21be and f0c2e32.

Files selected for processing: 5

• src/interpret.c
• src/pending_changes/gh-521-string-init-local-array-loop.txt
• tests/tests/regressions/.gh-521.rto/default.1.1.txt
• tests/tests/regressions/.gh-521.rto/default.2.1.txt
• tests/tests/regressions/gh-521.sft.cpp

Included review availability: This review used your included allowance. Your plan provides up to 4 included reviews per hour; 3 remain after this review.

[edgcpp-bot]

AcknowlEDG review results updated for f0c2e3214039:

• Bench - C (new)
• Bench - C++ (new)
• Test - Changes (new)

CI success.

[nv-cmeerw]

Looks good.