Description

Summary

This PR modifies the build.rs script in the libduckdb-sys crate to ensure deterministic builds by sorting the C++ file paths before including them in the build process.

Closes #582.

Motivation

Currently, cpp_files is derived from a HashSet, which yields non-deterministic ordering. This causes builds to vary across runs, leading to non-reproducible binaries. The change directly addresses the “Unreproducible binaries due to cpp_files HashSet in build.rs” issue by introducing a structured ordering of file inclusions.

Implementation Details

Convert the cpp_files iterator into a Vec.

Apply .sort() to that vector to enforce a consistent ordering.

Iterate over the sorted vector to invoke cfg.file(...) on each entry.

Test Results (Unchanged from Before)

These two tests continue to fail—they were failing prior to this PR and remain unrelated to its intent:

failures:

---- test_all_types::test_large_arrow_types stdout ----
Error: DuckDBFailure(Error { code: Unknown, extended_code: 1 }, Some("Binder Error: Column \"varint\" in EXCLUDE list not found in FROM clause"))

---- test_all_types::test_all_types stdout ----
Error: DuckDBFailure(Error { code: Unknown, extended_code: 1 }, Some("Binder Error: Column \"varint\" in EXCLUDE list not found in FROM clause"))

failures:
    test_all_types::test_all_types
    test_all_types::test_large_arrow_types

test result: FAILED. 123 passed; 2 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.88s

error: test failed, to rerun pass `-p duckdb --lib`

These errors are caused by a Binder Error—specifically stating that the column "varint" in the list was not found in the FROM clause—and are unrelated to the sorted C++ files.