[daveedvdv-nvidia]

Fixes #501.

During constant evaluation, the check that the number of elements allocated by an array new-expression suffices for its initializer took the length of a string literal initializer in bytes, rather than in characters. So, for a character type wider than one byte (e.g., wchar_t, char16_t, or char32_t), an array large enough for the literal could be diagnosed as too small:

constexpr int f(int n) {
  auto p = new wchar_t[n]{L"ab"};
  int r = p[1];
  delete[] p;
  return r;
}
static_assert(f(3) == L'b');  // Previously an error; now okay.

The interpreter now divides the byte length by the size of the element type. The new tests cover wchar_t, char16_t, char32_t, and char8_t (regressions/gh-501), and confirm that a too-short array is still diagnosed (regressions/gh-501-neg).

Summary by CodeRabbit

• Bug Fixes
  • Corrected compile-time evaluation of dynamically allocated arrays initialized with string literals. Array sizes are now checked in elements rather than bytes, avoiding incorrect size errors for wide-character strings.
  • Added regression coverage for wchar_t, char8_t, char16_t, and char32_t string initializers, including cases where the allocated array is too small.

[coderabbitai]

No actionable comments were generated in the recent review.

Walkthrough

The constant-evaluation array initializer check now compares the allocated element count with the string’s character count. C++20 regression tests cover wide string literals with sufficient and insufficient allocation sizes.

Changes: Constant-evaluation array initialization

Layer / File(s)	Summary
Count string initializer elements src/interpret.c	The initializer element count is now calculated by dividing the string’s byte length by the array element size.
Test string literal array sizes tests/tests/regressions/gh-501.sft.cpp, tests/tests/regressions/gh-501-neg.sft.cpp, tests/tests/regressions/.gh-501.rto/default.1.1.txt, tests/tests/regressions/.gh-501-neg.rto/default.1.1.txt, src/pending_changes/gh-501-array-new-wide-string.txt	C++20 regression tests cover accepted and rejected allocation sizes for wide string literals. The test fixtures record the invocation and expected diagnostic. The pending-changes note describes the issue and an example.

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

Change: Bug fix · Severity of issue fixed: Low

Suggested labels: bug

Suggested reviewers: wchilders-nvidia

Merge Risk: Minimal · up to 8c2d9

The change appears ready to merge after normal checks.

Pre-merge checks: 4

Check name	Status	Explanation
Description Check		Check skipped - CodeRabbit’s high-level summary is enabled.
Title check		The title clearly describes the main change: count string literal initializer characters rather than bytes during constant evaluation of array new-expressions.
Linked Issues check		Issue #501 requires constant evaluation to compare the allocated element count with the character count of a string-literal initializer. In src/interpret.c, do_constexpr_new now divides the stored…
Out of Scope Changes check		The implementation, regression tests, expected outputs, and pending-change note all address issue #501. The diff contains no demonstrated unrelated changes.

Full details: Linked Issues check

Explanation

Issue #501 requires constant evaluation to compare the allocated element count with the character count of a string-literal initializer. In src/interpret.c, do_constexpr_new now divides the stored byte length by the array element type size before the capacity check. tests/tests/regressions/gh-501.sft.cpp covers wchar_t, char16_t, char32_t, and char8_t; gh-501-neg.sft.cpp and its expected output confirm that a two-element wchar_t array remains too small for L"ab".

Finishing Touches

Generate docstrings

• Commit to this branch
• Create a new PR

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: 73b1b0f
• Layers: Repository YAML
• Review profile: CHILL
• Plan: Advanced
• Run ID: 7772e118-e872-4fed-86db-e33326197d5c

Commits

Reviewing files that changed from the base of the PR and between 73b1b0f and 8c2d918.

Files selected for processing: 6

• src/interpret.c
• src/pending_changes/gh-501-array-new-wide-string.txt
• tests/tests/regressions/.gh-501-neg.rto/default.1.1.txt
• tests/tests/regressions/.gh-501.rto/default.1.1.txt
• tests/tests/regressions/gh-501-neg.sft.cpp
• tests/tests/regressions/gh-501.sft.cpp

Included review availability: This review used your included allowance. Your plan provides up to 4 included reviews per hour; 2 remain after this review.

[edgcpp-bot]

AcknowlEDG review results updated for 8c2d918cd55d:

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

CI success.