fix(learn): generate starter files that are valid bash

[Chemaclass]

🤔 Background

Related #1256

bashunit learn writes a starter file per lesson, and 7 of the 10 templates did not parse — a function body of only TODO comments is a bash syntax error, and there were 28 of them:

functiontear_down() {
# TODO: Clean up test files
}

A learner who generated the file, filled in one TODO and ran the lesson got syntax error near unexpected token '}' from a file bashunit itself wrote — as their first contact with the tool.

💡 Changes

• A : no-op in each comment-only body, so every template parses
• run_lesson_test passes --fail-on-risky

The second is not optional. Making the templates valid alone is worse than the bug: an untouched template then completes the lesson, because the runner treats exit 0 as success and a test with no assertions is risky, which exits 0 by default.

template	run	result
comments only (before)	plain	syntax error
valid with :	plain	exit 0 — lesson completed, no work done
valid with :	--fail-on-risky	exit 1
solved	--fail-on-risky	exit 0

The learner now sees Test has no assertions (risky) — the actual state of their work.

Guard

Extracts every template and runs bash -n over it, plus a self-check that the scan can fail and an assertion that the runner keeps --fail-on-risky. Mutation-checked by removing one no-op.

My first extractor was line-oriented and reported templates as broken that bash -n accepts — it would have made the test lie. The perl one matches the measurement I had taken by hand, and the comment says why. ShellCheck cannot see inside a string literal, which is how 28 invalid bodies went unnoticed.

Verified through the real write path too: create_example_file uses echo "$content", and the templates contain '\'' escaping that arrives on disk correctly.

[Chemaclass]

🤔 Background

Related #1256

bashunit learn writes a starter file per lesson, and 7 of the 10 templates did not parse — a function body of only TODO comments is a bash syntax error, and there were 28 of them:

functiontear_down() {
# TODO: Clean up test files
}

A learner who generated the file, filled in one TODO and ran the lesson got syntax error near unexpected token '}' from a file bashunit itself wrote — as their first contact with the tool.

💡 Changes

• A : no-op in each comment-only body, so every template parses
• run_lesson_test passes --fail-on-risky

The second is not optional. Making the templates valid alone is worse than the bug: an untouched template then completes the lesson, because the runner treats exit 0 as success and a test with no assertions is risky, which exits 0 by default.

template	run	result
comments only (before)	plain	syntax error
valid with :	plain	exit 0 — lesson completed, no work done
valid with :	--fail-on-risky	exit 1
solved	--fail-on-risky	exit 0

The learner now sees Test has no assertions (risky) — the actual state of their work.

Guard

Extracts every template and runs bash -n over it, plus a self-check that the scan can fail and an assertion that the runner keeps --fail-on-risky. Mutation-checked by removing one no-op.

My first extractor was line-oriented and reported templates as broken that bash -n accepts — it would have made the test lie. The perl one matches the measurement I had taken by hand, and the comment says why. ShellCheck cannot see inside a string literal, which is how 28 invalid bodies went unnoticed.

Verified through the real write path too: create_example_file uses echo "$content", and the templates contain '\'' escaping that arrives on disk correctly.

[Chemaclass]

🤔 Background

Related #1256

bashunit learn writes a starter file per lesson, and 7 of the 10 templates did not parse — a function body of only TODO comments is a bash syntax error, and there were 28 of them:

functiontear_down() {
# TODO: Clean up test files
}

A learner who generated the file, filled in one TODO and ran the lesson got syntax error near unexpected token '}' from a file bashunit itself wrote — as their first contact with the tool.

💡 Changes

• A : no-op in each comment-only body, so every template parses
• run_lesson_test passes --fail-on-risky

The second is not optional. Making the templates valid alone is worse than the bug: an untouched template then completes the lesson, because the runner treats exit 0 as success and a test with no assertions is risky, which exits 0 by default.

template	run	result
comments only (before)	plain	syntax error
valid with :	plain	exit 0 — lesson completed, no work done
valid with :	--fail-on-risky	exit 1
solved	--fail-on-risky	exit 0

The learner now sees Test has no assertions (risky) — the actual state of their work.

Guard

Extracts every template and runs bash -n over it, plus a self-check that the scan can fail and an assertion that the runner keeps --fail-on-risky. Mutation-checked by removing one no-op.

My first extractor was line-oriented and reported templates as broken that bash -n accepts — it would have made the test lie. The perl one matches the measurement I had taken by hand, and the comment says why. ShellCheck cannot see inside a string literal, which is how 28 invalid bodies went unnoticed.

Verified through the real write path too: create_example_file uses echo "$content", and the templates contain '\'' escaping that arrives on disk correctly.

[Chemaclass]

🤔 Background

Related #1256

bashunit learn writes a starter file per lesson, and 7 of the 10 templates did not parse — a function body of only TODO comments is a bash syntax error, and there were 28 of them:

functiontear_down() {
# TODO: Clean up test files
}

A learner who generated the file, filled in one TODO and ran the lesson got syntax error near unexpected token '}' from a file bashunit itself wrote — as their first contact with the tool.

💡 Changes

• A : no-op in each comment-only body, so every template parses
• run_lesson_test passes --fail-on-risky

The second is not optional. Making the templates valid alone is worse than the bug: an untouched template then completes the lesson, because the runner treats exit 0 as success and a test with no assertions is risky, which exits 0 by default.

template	run	result
comments only (before)	plain	syntax error
valid with :	plain	exit 0 — lesson completed, no work done
valid with :	--fail-on-risky	exit 1
solved	--fail-on-risky	exit 0

The learner now sees Test has no assertions (risky) — the actual state of their work.

Guard

Extracts every template and runs bash -n over it, plus a self-check that the scan can fail and an assertion that the runner keeps --fail-on-risky. Mutation-checked by removing one no-op.

My first extractor was line-oriented and reported templates as broken that bash -n accepts — it would have made the test lie. The perl one matches the measurement I had taken by hand, and the comment says why. ShellCheck cannot see inside a string literal, which is how 28 invalid bodies went unnoticed.

Verified through the real write path too: create_example_file uses echo "$content", and the templates contain '\'' escaping that arrives on disk correctly.

[Chemaclass]

🤔 Background

Related #1256

bashunit learn writes a starter file per lesson, and 7 of the 10 templates did not parse — a function body of only TODO comments is a bash syntax error, and there were 28 of them:

functiontear_down() {
# TODO: Clean up test files
}

A learner who generated the file, filled in one TODO and ran the lesson got syntax error near unexpected token '}' from a file bashunit itself wrote — as their first contact with the tool.

💡 Changes

• A : no-op in each comment-only body, so every template parses
• run_lesson_test passes --fail-on-risky

The second is not optional. Making the templates valid alone is worse than the bug: an untouched template then completes the lesson, because the runner treats exit 0 as success and a test with no assertions is risky, which exits 0 by default.

template	run	result
comments only (before)	plain	syntax error
valid with :	plain	exit 0 — lesson completed, no work done
valid with :	--fail-on-risky	exit 1
solved	--fail-on-risky	exit 0

The learner now sees Test has no assertions (risky) — the actual state of their work.

Guard

Extracts every template and runs bash -n over it, plus a self-check that the scan can fail and an assertion that the runner keeps --fail-on-risky. Mutation-checked by removing one no-op.

My first extractor was line-oriented and reported templates as broken that bash -n accepts — it would have made the test lie. The perl one matches the measurement I had taken by hand, and the comment says why. ShellCheck cannot see inside a string literal, which is how 28 invalid bodies went unnoticed.

Verified through the real write path too: create_example_file uses echo "$content", and the templates contain '\'' escaping that arrives on disk correctly.

[Chemaclass]

🤔 Background

Related #1256

bashunit learn writes a starter file per lesson, and 7 of the 10 templates did not parse — a function body of only TODO comments is a bash syntax error, and there were 28 of them:

functiontear_down() {
# TODO: Clean up test files
}

A learner who generated the file, filled in one TODO and ran the lesson got syntax error near unexpected token '}' from a file bashunit itself wrote — as their first contact with the tool.

💡 Changes

• A : no-op in each comment-only body, so every template parses
• run_lesson_test passes --fail-on-risky

The second is not optional. Making the templates valid alone is worse than the bug: an untouched template then completes the lesson, because the runner treats exit 0 as success and a test with no assertions is risky, which exits 0 by default.

template	run	result
comments only (before)	plain	syntax error
valid with :	plain	exit 0 — lesson completed, no work done
valid with :	--fail-on-risky	exit 1
solved	--fail-on-risky	exit 0

The learner now sees Test has no assertions (risky) — the actual state of their work.

Guard

Extracts every template and runs bash -n over it, plus a self-check that the scan can fail and an assertion that the runner keeps --fail-on-risky. Mutation-checked by removing one no-op.

My first extractor was line-oriented and reported templates as broken that bash -n accepts — it would have made the test lie. The perl one matches the measurement I had taken by hand, and the comment says why. ShellCheck cannot see inside a string literal, which is how 28 invalid bodies went unnoticed.

Verified through the real write path too: create_example_file uses echo "$content", and the templates contain '\'' escaping that arrives on disk correctly.

[Chemaclass]

🤔 Background

Related #1256

bashunit learn writes a starter file per lesson, and 7 of the 10 templates did not parse — a function body of only TODO comments is a bash syntax error, and there were 28 of them:

functiontear_down() {
# TODO: Clean up test files
}

A learner who generated the file, filled in one TODO and ran the lesson got syntax error near unexpected token '}' from a file bashunit itself wrote — as their first contact with the tool.

💡 Changes

• A : no-op in each comment-only body, so every template parses
• run_lesson_test passes --fail-on-risky

The second is not optional. Making the templates valid alone is worse than the bug: an untouched template then completes the lesson, because the runner treats exit 0 as success and a test with no assertions is risky, which exits 0 by default.

template	run	result
comments only (before)	plain	syntax error
valid with :	plain	exit 0 — lesson completed, no work done
valid with :	--fail-on-risky	exit 1
solved	--fail-on-risky	exit 0

The learner now sees Test has no assertions (risky) — the actual state of their work.

Guard

Extracts every template and runs bash -n over it, plus a self-check that the scan can fail and an assertion that the runner keeps --fail-on-risky. Mutation-checked by removing one no-op.

My first extractor was line-oriented and reported templates as broken that bash -n accepts — it would have made the test lie. The perl one matches the measurement I had taken by hand, and the comment says why. ShellCheck cannot see inside a string literal, which is how 28 invalid bodies went unnoticed.

Verified through the real write path too: create_example_file uses echo "$content", and the templates contain '\'' escaping that arrives on disk correctly.

[Chemaclass]

🤔 Background

Related #1256

bashunit learn writes a starter file per lesson, and 7 of the 10 templates did not parse — a function body of only TODO comments is a bash syntax error, and there were 28 of them:

functiontear_down() {
# TODO: Clean up test files
}

A learner who generated the file, filled in one TODO and ran the lesson got syntax error near unexpected token '}' from a file bashunit itself wrote — as their first contact with the tool.

💡 Changes

• A : no-op in each comment-only body, so every template parses
• run_lesson_test passes --fail-on-risky

The second is not optional. Making the templates valid alone is worse than the bug: an untouched template then completes the lesson, because the runner treats exit 0 as success and a test with no assertions is risky, which exits 0 by default.

template	run	result
comments only (before)	plain	syntax error
valid with :	plain	exit 0 — lesson completed, no work done
valid with :	--fail-on-risky	exit 1
solved	--fail-on-risky	exit 0

The learner now sees Test has no assertions (risky) — the actual state of their work.

Guard

Extracts every template and runs bash -n over it, plus a self-check that the scan can fail and an assertion that the runner keeps --fail-on-risky. Mutation-checked by removing one no-op.

My first extractor was line-oriented and reported templates as broken that bash -n accepts — it would have made the test lie. The perl one matches the measurement I had taken by hand, and the comment says why. ShellCheck cannot see inside a string literal, which is how 28 invalid bodies went unnoticed.

Verified through the real write path too: create_example_file uses echo "$content", and the templates contain '\'' escaping that arrives on disk correctly.