Hmbown/CodeWhale · warning · anyhow::Error

No diff to review.

Error message

No diff to review.

What it means

run_review() first collects the diff via collect_diff() — plain `git diff`, plus `--cached` with --staged, `base...HEAD` with --base, and an optional `--path` filter — and bails when the result trims to empty. There is nothing for the review model to look at, so the command aborts before model/route resolution.

Source

Thrown at crates/tui/src/lib.rs:7933

    let input = input.trim();
    if input.is_empty() {
        bail!("No session selected.");
    }
    let idx: usize = input
        .parse()
        .map_err(|_| anyhow::anyhow!("Invalid input"))?;
    let session = sessions
        .get(idx.saturating_sub(1))
        .ok_or_else(|| anyhow::anyhow!("Selection out of range"))?;
    Ok(session.id.clone())
}

async fn run_review(config: &Config, args: ReviewArgs) -> Result<()> {
    use crate::client::DeepSeekClient;

    let diff = collect_diff(&args)?;
    if diff.trim().is_empty() {
        bail!("No diff to review.");
    }
    validate_review_receipt_args(&args)?;
    if args.check_receipt {
        return run_review_receipt_check(&diff, &args);
    }

    let model = resolve_review_model(config, args.model.as_deref());
    let route = resolve_cli_exec_route(config, &model, &diff, args.model.is_none()).await?;
    let execution_config = config_for_cli_route(config, &route);
    let route_provider = execution_config.provider_identity_for(route.provider);
    let model = route.model.clone();
    let user_prompt =
        format!("Review the following diff and provide feedback:\n\n{diff}\n\nEnd of diff.");
    let reasoning_effort = route.reasoning_effort.and_then(|effort| {
        cli_reasoning_effort_value_for_prompt(&execution_config, &model, effort, &user_prompt)
    });

    let system = SystemPrompt::Text(

View on GitHub (pinned to 0c42157ee5)

Solutions

1. Unstaged changes: ensure the tree differs from the index, or pass `--staged` when changes are staged
2. Review a committed change with `--base` (e.g. `--base HEAD~1`)
3. Check the `--path` filter matches files that actually changed (`git status`)
4. Confirm you are inside the intended git repository

Example fix

// before: everything already committed, tree is clean
$ git commit -am wip && codewhale review
Error: No diff to review.

// after: review the just-committed change against its parent
$ codewhale review --base HEAD~1

# mirror collect_diff's selection before invoking codewhale review
if git diff --quiet && git diff --cached --quiet; then
  echo 'nothing to review: clean tree, nothing staged' >&2; exit 2
fi
codewhale review

When it happens

Trigger: Clean working tree with no --staged/--base (default diff empty); `--staged` with nothing staged; a `--base` equal to HEAD; or a `--path` filter matching no changes.

Common situations: Committing everything just before reviewing; forgetting --staged when all changes are staged; reviewing in the wrong repo/directory; a mistyped pathspec.

Related errors

• git diff failed: {}
• empty API key provided
• Model name cannot be empty
• worktree branch must not be empty
• repo root does not exist: {}

AI-assisted analysis of Hmbown/CodeWhale@0c42157ee5 (2026-08-20). Data as JSON: /api/errors/b23369056e660154. Report an issue: GitHub.