perf(runner): read failing-test body once for Source: context by Chemaclass · Pull Request #807 · TypedDevs/bashunit

Chemaclass · 2026-07-15T22:29:01Z

🤔 Background

Profiling a failing test surfaced quadratic fork behaviour: get_failure_source_context re-read the test file with a sed -n Np fork per line and a grep fork per line to find the function's closing brace, so the fork count grew with the function length.

💡 Changes

Read the function body in a single bash read loop and detect the closing brace by trimming whitespace — no per-line sed/grep forks (a 20-line body dropped from ~46 to ~2 forks).
Output is byte-identical; failure snapshots unchanged; regression test added. No behaviour change.

get_failure_source_context re-read the test file with a 'sed -n Np' fork for every line of the function body and a 'grep' fork to detect the closing brace, so fork count grew with function length (a 20-line body cost ~46 forks). Read the body in a single bash 'read' loop and detect the closing brace by trimming whitespace — no per-line forks (~46 -> ~2). Output is byte-identical (failure snapshots unchanged). No behaviour change.

Chemaclass · 2026-07-15T22:29:01Z

🤔 Background

Profiling a failing test surfaced quadratic fork behaviour: get_failure_source_context re-read the test file with a sed -n Np fork per line and a grep fork per line to find the function's closing brace, so the fork count grew with the function length.

💡 Changes

Read the function body in a single bash read loop and detect the closing brace by trimming whitespace — no per-line sed/grep forks (a 20-line body dropped from ~46 to ~2 forks).
Output is byte-identical; failure snapshots unchanged; regression test added. No behaviour change.

get_failure_source_context re-read the test file with a 'sed -n Np' fork for every line of the function body and a 'grep' fork to detect the closing brace, so fork count grew with function length (a 20-line body cost ~46 forks). Read the body in a single bash 'read' loop and detect the closing brace by trimming whitespace — no per-line forks (~46 -> ~2). Output is byte-identical (failure snapshots unchanged). No behaviour change.

Chemaclass · 2026-07-15T22:29:01Z

🤔 Background

Profiling a failing test surfaced quadratic fork behaviour: get_failure_source_context re-read the test file with a sed -n Np fork per line and a grep fork per line to find the function's closing brace, so the fork count grew with the function length.

💡 Changes

Read the function body in a single bash read loop and detect the closing brace by trimming whitespace — no per-line sed/grep forks (a 20-line body dropped from ~46 to ~2 forks).
Output is byte-identical; failure snapshots unchanged; regression test added. No behaviour change.

get_failure_source_context re-read the test file with a 'sed -n Np' fork for every line of the function body and a 'grep' fork to detect the closing brace, so fork count grew with function length (a 20-line body cost ~46 forks). Read the body in a single bash 'read' loop and detect the closing brace by trimming whitespace — no per-line forks (~46 -> ~2). Output is byte-identical (failure snapshots unchanged). No behaviour change.

Chemaclass · 2026-07-15T22:29:01Z

🤔 Background

Profiling a failing test surfaced quadratic fork behaviour: get_failure_source_context re-read the test file with a sed -n Np fork per line and a grep fork per line to find the function's closing brace, so the fork count grew with the function length.

💡 Changes

Read the function body in a single bash read loop and detect the closing brace by trimming whitespace — no per-line sed/grep forks (a 20-line body dropped from ~46 to ~2 forks).
Output is byte-identical; failure snapshots unchanged; regression test added. No behaviour change.

get_failure_source_context re-read the test file with a 'sed -n Np' fork for every line of the function body and a 'grep' fork to detect the closing brace, so fork count grew with function length (a 20-line body cost ~46 forks). Read the body in a single bash 'read' loop and detect the closing brace by trimming whitespace — no per-line forks (~46 -> ~2). Output is byte-identical (failure snapshots unchanged). No behaviour change.

Chemaclass · 2026-07-15T22:29:01Z

🤔 Background

Profiling a failing test surfaced quadratic fork behaviour: get_failure_source_context re-read the test file with a sed -n Np fork per line and a grep fork per line to find the function's closing brace, so the fork count grew with the function length.

💡 Changes

Read the function body in a single bash read loop and detect the closing brace by trimming whitespace — no per-line sed/grep forks (a 20-line body dropped from ~46 to ~2 forks).
Output is byte-identical; failure snapshots unchanged; regression test added. No behaviour change.

get_failure_source_context re-read the test file with a 'sed -n Np' fork for every line of the function body and a 'grep' fork to detect the closing brace, so fork count grew with function length (a 20-line body cost ~46 forks). Read the body in a single bash 'read' loop and detect the closing brace by trimming whitespace — no per-line forks (~46 -> ~2). Output is byte-identical (failure snapshots unchanged). No behaviour change.

Chemaclass · 2026-07-15T22:29:01Z

🤔 Background

Profiling a failing test surfaced quadratic fork behaviour: get_failure_source_context re-read the test file with a sed -n Np fork per line and a grep fork per line to find the function's closing brace, so the fork count grew with the function length.

💡 Changes

Read the function body in a single bash read loop and detect the closing brace by trimming whitespace — no per-line sed/grep forks (a 20-line body dropped from ~46 to ~2 forks).
Output is byte-identical; failure snapshots unchanged; regression test added. No behaviour change.

get_failure_source_context re-read the test file with a 'sed -n Np' fork for every line of the function body and a 'grep' fork to detect the closing brace, so fork count grew with function length (a 20-line body cost ~46 forks). Read the body in a single bash 'read' loop and detect the closing brace by trimming whitespace — no per-line forks (~46 -> ~2). Output is byte-identical (failure snapshots unchanged). No behaviour change.

Chemaclass · 2026-07-15T22:29:01Z

🤔 Background

Profiling a failing test surfaced quadratic fork behaviour: get_failure_source_context re-read the test file with a sed -n Np fork per line and a grep fork per line to find the function's closing brace, so the fork count grew with the function length.

💡 Changes

Read the function body in a single bash read loop and detect the closing brace by trimming whitespace — no per-line sed/grep forks (a 20-line body dropped from ~46 to ~2 forks).
Output is byte-identical; failure snapshots unchanged; regression test added. No behaviour change.

get_failure_source_context re-read the test file with a 'sed -n Np' fork for every line of the function body and a 'grep' fork to detect the closing brace, so fork count grew with function length (a 20-line body cost ~46 forks). Read the body in a single bash 'read' loop and detect the closing brace by trimming whitespace — no per-line forks (~46 -> ~2). Output is byte-identical (failure snapshots unchanged). No behaviour change.

Chemaclass · 2026-07-15T22:29:01Z

🤔 Background

Profiling a failing test surfaced quadratic fork behaviour: get_failure_source_context re-read the test file with a sed -n Np fork per line and a grep fork per line to find the function's closing brace, so the fork count grew with the function length.

💡 Changes

Read the function body in a single bash read loop and detect the closing brace by trimming whitespace — no per-line sed/grep forks (a 20-line body dropped from ~46 to ~2 forks).
Output is byte-identical; failure snapshots unchanged; regression test added. No behaviour change.

get_failure_source_context re-read the test file with a 'sed -n Np' fork for every line of the function body and a 'grep' fork to detect the closing brace, so fork count grew with function length (a 20-line body cost ~46 forks). Read the body in a single bash 'read' loop and detect the closing brace by trimming whitespace — no per-line forks (~46 -> ~2). Output is byte-identical (failure snapshots unchanged). No behaviour change.