Fix: cap aggregated exec output consistently

[Kbediako]

WHAT?

• Bias aggregated output toward stderr under contention (2/3 stderr, 1/3 stdout) while keeping the 1 MiB cap.
• Rebalance unused stderr share back to stdout when stderr is tiny to avoid underfilling.
• Add tests for contention, small-stderr rebalance, and under-cap ordering (stdout then stderr).

WHY?

• Review feedback requested stderr priority under contention.
• Avoid underfilled aggregated output when stderr is small while preserving a consistent cap across exec paths.

HOW?

• Update aggregate_output to compute stdout/stderr shares, then reassign unused capacity to the other stream.
• Use the helper in both Windows and async exec paths.
• Add regression tests for contention/rebalance and under-cap ordering.

BEFORE

// Best-effort aggregate: stdout then stderr (capped).
let mut aggregated = Vec::with_capacity(
    stdout
        .text
        .len()
        .saturating_add(stderr.text.len())
        .min(EXEC_OUTPUT_MAX_BYTES),
);
append_capped(&mut aggregated, &stdout.text, EXEC_OUTPUT_MAX_BYTES);
append_capped(&mut aggregated, &stderr.text, EXEC_OUTPUT_MAX_BYTES);
let aggregated_output = StreamOutput {
    text: aggregated,
    truncated_after_lines: None,
};

AFTER

fn aggregate_output(
    stdout: &StreamOutput<Vec<u8>>,
    stderr: &StreamOutput<Vec<u8>>,
) -> StreamOutput<Vec<u8>> {
    let total_len = stdout.text.len().saturating_add(stderr.text.len());
    let max_bytes = EXEC_OUTPUT_MAX_BYTES;
    let mut aggregated = Vec::with_capacity(total_len.min(max_bytes));

    if total_len <= max_bytes {
        aggregated.extend_from_slice(&stdout.text);
        aggregated.extend_from_slice(&stderr.text);
        return StreamOutput {
            text: aggregated,
            truncated_after_lines: None,
        };
    }

    // Under contention, reserve 1/3 for stdout and 2/3 for stderr; rebalance unused stderr to stdout.
    let want_stdout = stdout.text.len().min(max_bytes / 3);
    let want_stderr = stderr.text.len();
    let stderr_take = want_stderr.min(max_bytes.saturating_sub(want_stdout));
    let remaining = max_bytes.saturating_sub(want_stdout + stderr_take);
    let stdout_take = want_stdout + remaining.min(stdout.text.len().saturating_sub(want_stdout));

    aggregated.extend_from_slice(&stdout.text[..stdout_take]);
    aggregated.extend_from_slice(&stderr.text[..stderr_take]);

    StreamOutput {
        text: aggregated,
        truncated_after_lines: None,
    }
}

TESTS

• just fmt
• just fix -p codex-core
• cargo test -p codex-core aggregate_output_
• cargo test -p codex-core
• cargo test --all-features

FIXES

Fixes #9758

[Kbediako]

Note: 6 "Local Bazel build" checks are failing. The logs show BuildBuddy cache auth errors (UNAUTHENTICATED: missing API key), which is typical for forked PRs; not a code failure on this change.

[jif-oai]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Can't wait for the next one!

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[jif-oai]

lgtm

[jif-oai]

Just FYI: ok for this because in bytes but using [] accessor in Rust in generally dangerous as it can panic. For example if you have a string containing multi-bytes characters, this accessor can split a char between its bytes and it will panic the runtime