fix(lib): reload config on watch restart

[fengmk2]

Pass the build function as the restart callback to buildWithConfigs
instead of an empty function, ensuring config is re-resolved when
watch mode detects config file changes.

Ref: rolldown/tsdown@1756b03

close VP-160

[fengmk2]

• test: double snap-test concurrency limit #507
• fix(lib): reload config on watch restart #506 👈 (View in Graphite)
• main

This stack of pull requests is managed by Graphite. Learn more about stacking.

[linear]

VP-160 follow tsdown bugfix: reload config on restart

rolldown/tsdown@1756b03

[fengmk2]

rolldown watch mode is not working on Github Action CI env, but I run this snap-test on local it was worked. Skip it for now.

[fengmk2]

Merge activity

• Jan 26, 6:30 AM UTC: A user started a stack merge that includes this pull request via Graphite.
• Jan 26, 6:30 AM UTC: Graphite rebased this pull request as part of a merge.
• Jan 26, 6:37 AM UTC: @fengmk2 merged this pull request with Graphite.

[Copilot]

Pull request overview

Updates the vite lib CLI watch restart behavior so that config is re-resolved when a watched config file change triggers a restart, and extends the snapshot test runner to support per-command timeouts and always-run cleanup commands.

Changes:

• Pass the build function as the watch-restart callback to buildWithConfigs so restarts re-resolve config.
• Add timeout per command and after cleanup commands to the snap-test runner.
• Add a (currently todo) snap-test fixture for verifying config reload/restart behavior in watch mode.

Reviewed changes

Copilot reviewed 11 out of 11 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
packages/tools/src/snap-test.ts	Adds Command.timeout and Steps.after, and runs after commands in a finally block.
packages/cli/src/lib-bin.ts	Wraps build in runBuild() and passes it as the restart callback to re-resolve config on watch restarts.
packages/cli/snap-tests-todo/command-lib-watch-restart/wait-for-dist2.sh	Adds helper script to wait for dist2 after config change.
packages/cli/snap-tests-todo/command-lib-watch-restart/wait-for-dist.sh	Adds helper script to wait for initial dist.
packages/cli/snap-tests-todo/command-lib-watch-restart/vite.config.ts	Adds minimal config used by the watch-restart snapshot fixture.
packages/cli/snap-tests-todo/command-lib-watch-restart/steps.json	Defines the watch-restart scenario, including a longer timeout and cleanup via after.
packages/cli/snap-tests-todo/command-lib-watch-restart/src/index.ts	Adds minimal source input for the fixture.
packages/cli/snap-tests-todo/command-lib-watch-restart/snap.txt	Adds expected snapshot output for the fixture.
packages/cli/snap-tests-todo/command-lib-watch-restart/run-watch.sh	Starts vite lib --watch in background and writes PID/log file.
packages/cli/snap-tests-todo/command-lib-watch-restart/package.json	Declares the fixture package metadata.
packages/cli/snap-tests-todo/command-lib-watch-restart/kill-watch.sh	Adds cleanup script to stop the background watch process tree.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

outputStream is only closed on the success path. If execute(...) rejects (e.g., command not found / spawn error) or Promise.race rejects, the function will jump to the outer finally and leak the open file descriptor for output.log. Consider wrapping each command execution in a per-command try/finally that always closes outputStream (and ideally still reads any partial output).

Suggested change

	const outputStream = await open(outputStreamPath, 'w');
	const exitCode = await Promise.race([
	execute(stripComments(cmd.command), [], {
	env,
	cwd,
	stdin: null,
	// Declared to be `Writable` but `FileHandle` works too.
	// @ts-expect-error
	stderr: outputStream,
	// @ts-expect-error
	stdout: outputStream,
	glob: {
	// Disable glob expansion. Pass args like '--filter=*' as-is.
	isGlobPattern: () => false,
	match: async () => [],
	},
	}),
	setTimeout(cmd.timeout ?? 30 * 1000),
	]);
	await outputStream.close();
	let output = readFileSync(outputStreamPath, 'utf-8');
	let commandLine = `> ${cmd.command}`;
	if (exitCode !== 0) {
	commandLine = (exitCode === undefined ? '[timeout]' : `[${exitCode}]`) + commandLine;
	} else {
	// only allow ignore output if the command is successful
	if (cmd.ignoreOutput) {
	output = '';
	}
	}
	newSnap.push(commandLine);
	if (output.length > 0) {
	newSnap.push(replaceUnstableOutput(output, caseTmpDir));
	}
	if (exitCode === undefined) {
	break; // Stop executing further commands on timeout
	}
	let outputStream: fsPromises.FileHandle | null = null;
	try {
	outputStream = await open(outputStreamPath, 'w');
	const exitCode = await Promise.race([
	execute(stripComments(cmd.command), [], {
	env,
	cwd,
	stdin: null,
	// Declared to be `Writable` but `FileHandle` works too.
	// @ts-expect-error
	stderr: outputStream,
	// @ts-expect-error
	stdout: outputStream,
	glob: {
	// Disable glob expansion. Pass args like '--filter=*' as-is.
	isGlobPattern: () => false,
	match: async () => [],
	},
	}),
	setTimeout(cmd.timeout ?? 30 * 1000),
	]);
	await outputStream.close();
	outputStream = null;
	let output = readFileSync(outputStreamPath, 'utf-8');
	let commandLine = `> ${cmd.command}`;
	if (exitCode !== 0) {
	commandLine = (exitCode === undefined ? '[timeout]' : `[${exitCode}]`) + commandLine;
	} else {
	// only allow ignore output if the command is successful
	if (cmd.ignoreOutput) {
	output = '';
	}
	}
	newSnap.push(commandLine);
	if (output.length > 0) {
	newSnap.push(replaceUnstableOutput(output, caseTmpDir));
	}
	if (exitCode === undefined) {
	break; // Stop executing further commands on timeout
	}
	} finally {
	if (outputStream !== null) {
	try {
	await outputStream.close();
	} catch {
	// Ignore errors during cleanup close
	}
	}
	}

[Copilot]

The PR description focuses on reloading config on watch restart, but this change also expands the snap-test JSON schema (Command.timeout, Steps.after) and adds new test fixtures. If this broader scope is intentional, it would help to either mention it in the PR description or split the test-runner/schema changes into a separate PR for easier review and rollout.