chore(sa): remove 11 dead ShellCheck disable directives by Chemaclass · Pull Request #963 · TypedDevs/bashunit

Chemaclass · 2026-08-02T09:06:46Z

🤔 Background

A # shellcheck disable= for a warning that no longer fires is dead weight — and misleading, since it tells the next reader the line below is doing something ShellCheck objects to when it is not. This project has found stale ones twice before (#928, #947), both times by accident rather than by looking.

💡 Changes

Removed 11 dead directives across 9 files: SC2153, SC2034 ×3, SC2154 ×3, SC2254, SC2030, SC2031, SC2295
Kept every SC1091 / SC2155 / SC2016 disable — those don't fire only because both gates exclude those codes globally, which is not the same as being obsolete; they still work for anyone running plain shellcheck
Explanatory comments above or beside a removed directive are kept where they still say something true

🔍 How they were found

Mechanically, not by eye: strip every disable from a file, re-run ShellCheck, see which declared codes come back. Two details decided correct-vs-wrong:

In place, not in /tmp..shellcheckrc sets source-path=src + external-sources=true, so copying a file elsewhere breaks source resolution and changes what SC2154/SC2034 report — the exact codes under test.
Both gates.make sa runs -xC (follows source); CI's ludeeus/action-shellcheck does not. They disagree on cross-file symbols, so a directive only counts as dead if it fires under neither. That didn't change the final list, but assuming it would have been luck.

✅ Verification

Every changed line is a comment (diff filtered to confirm). make sa · make lint · CI-mode ShellCheck (no -x, exact CI opts, clean) · bash build.sh bin -v → ✅ Build verified ✅.

Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

A directive for a warning that no longer fires is dead weight, and worse than that it is misleading -- it tells the next reader that the line below is doing something ShellCheck objects to, when it is not. This project has found stale ones twice before, in #928 and #947, both times by accident. Detected mechanically rather than by eye: strip every disable from a file, run ShellCheck, and see which declared codes actually come back. Two details made the difference between a correct list and a wrong one. The scan has to run in place. `.shellcheckrc` sets `source-path=src` with `external-sources=true`, so copying a file to a temp dir breaks source resolution and changes what SC2154 and SC2034 report -- exactly the codes under test. And it has to run under both gates. `make sa` uses `-xC` (follows `source`), CI's ludeeus/action-shellcheck does not, and the two see different things for cross-file symbols. A directive only counts as dead if it fires under neither. That distinction turned out not to change the list here, but assuming it would have been luck rather than method. Kept: the disables for SC1091, SC2155 and SC2016. Those never fire because both gates already exclude those codes globally, which is not the same as the warning being obsolete -- they still do their job for anyone running plain `shellcheck`. Removed (9 files, 11 lines): SC2153, SC2034 x3, SC2154 x3, SC2254, SC2030, SC2031, SC2295. Explanatory comments that sat above or beside a removed directive are kept where they still say something true. Every changed line is a comment; verified by filtering the diff. Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

Chemaclass · 2026-08-02T09:06:46Z

🤔 Background

A # shellcheck disable= for a warning that no longer fires is dead weight — and misleading, since it tells the next reader the line below is doing something ShellCheck objects to when it is not. This project has found stale ones twice before (#928, #947), both times by accident rather than by looking.

💡 Changes

Removed 11 dead directives across 9 files: SC2153, SC2034 ×3, SC2154 ×3, SC2254, SC2030, SC2031, SC2295
Kept every SC1091 / SC2155 / SC2016 disable — those don't fire only because both gates exclude those codes globally, which is not the same as being obsolete; they still work for anyone running plain shellcheck
Explanatory comments above or beside a removed directive are kept where they still say something true

🔍 How they were found

Mechanically, not by eye: strip every disable from a file, re-run ShellCheck, see which declared codes come back. Two details decided correct-vs-wrong:

In place, not in /tmp..shellcheckrc sets source-path=src + external-sources=true, so copying a file elsewhere breaks source resolution and changes what SC2154/SC2034 report — the exact codes under test.
Both gates.make sa runs -xC (follows source); CI's ludeeus/action-shellcheck does not. They disagree on cross-file symbols, so a directive only counts as dead if it fires under neither. That didn't change the final list, but assuming it would have been luck.

✅ Verification

Every changed line is a comment (diff filtered to confirm). make sa · make lint · CI-mode ShellCheck (no -x, exact CI opts, clean) · bash build.sh bin -v → ✅ Build verified ✅.

Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

A directive for a warning that no longer fires is dead weight, and worse than that it is misleading -- it tells the next reader that the line below is doing something ShellCheck objects to, when it is not. This project has found stale ones twice before, in #928 and #947, both times by accident. Detected mechanically rather than by eye: strip every disable from a file, run ShellCheck, and see which declared codes actually come back. Two details made the difference between a correct list and a wrong one. The scan has to run in place. `.shellcheckrc` sets `source-path=src` with `external-sources=true`, so copying a file to a temp dir breaks source resolution and changes what SC2154 and SC2034 report -- exactly the codes under test. And it has to run under both gates. `make sa` uses `-xC` (follows `source`), CI's ludeeus/action-shellcheck does not, and the two see different things for cross-file symbols. A directive only counts as dead if it fires under neither. That distinction turned out not to change the list here, but assuming it would have been luck rather than method. Kept: the disables for SC1091, SC2155 and SC2016. Those never fire because both gates already exclude those codes globally, which is not the same as the warning being obsolete -- they still do their job for anyone running plain `shellcheck`. Removed (9 files, 11 lines): SC2153, SC2034 x3, SC2154 x3, SC2254, SC2030, SC2031, SC2295. Explanatory comments that sat above or beside a removed directive are kept where they still say something true. Every changed line is a comment; verified by filtering the diff. Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

Chemaclass · 2026-08-02T09:06:46Z

🤔 Background

A # shellcheck disable= for a warning that no longer fires is dead weight — and misleading, since it tells the next reader the line below is doing something ShellCheck objects to when it is not. This project has found stale ones twice before (#928, #947), both times by accident rather than by looking.

💡 Changes

Removed 11 dead directives across 9 files: SC2153, SC2034 ×3, SC2154 ×3, SC2254, SC2030, SC2031, SC2295
Kept every SC1091 / SC2155 / SC2016 disable — those don't fire only because both gates exclude those codes globally, which is not the same as being obsolete; they still work for anyone running plain shellcheck
Explanatory comments above or beside a removed directive are kept where they still say something true

🔍 How they were found

Mechanically, not by eye: strip every disable from a file, re-run ShellCheck, see which declared codes come back. Two details decided correct-vs-wrong:

In place, not in /tmp..shellcheckrc sets source-path=src + external-sources=true, so copying a file elsewhere breaks source resolution and changes what SC2154/SC2034 report — the exact codes under test.
Both gates.make sa runs -xC (follows source); CI's ludeeus/action-shellcheck does not. They disagree on cross-file symbols, so a directive only counts as dead if it fires under neither. That didn't change the final list, but assuming it would have been luck.

✅ Verification

Every changed line is a comment (diff filtered to confirm). make sa · make lint · CI-mode ShellCheck (no -x, exact CI opts, clean) · bash build.sh bin -v → ✅ Build verified ✅.

Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

A directive for a warning that no longer fires is dead weight, and worse than that it is misleading -- it tells the next reader that the line below is doing something ShellCheck objects to, when it is not. This project has found stale ones twice before, in #928 and #947, both times by accident. Detected mechanically rather than by eye: strip every disable from a file, run ShellCheck, and see which declared codes actually come back. Two details made the difference between a correct list and a wrong one. The scan has to run in place. `.shellcheckrc` sets `source-path=src` with `external-sources=true`, so copying a file to a temp dir breaks source resolution and changes what SC2154 and SC2034 report -- exactly the codes under test. And it has to run under both gates. `make sa` uses `-xC` (follows `source`), CI's ludeeus/action-shellcheck does not, and the two see different things for cross-file symbols. A directive only counts as dead if it fires under neither. That distinction turned out not to change the list here, but assuming it would have been luck rather than method. Kept: the disables for SC1091, SC2155 and SC2016. Those never fire because both gates already exclude those codes globally, which is not the same as the warning being obsolete -- they still do their job for anyone running plain `shellcheck`. Removed (9 files, 11 lines): SC2153, SC2034 x3, SC2154 x3, SC2254, SC2030, SC2031, SC2295. Explanatory comments that sat above or beside a removed directive are kept where they still say something true. Every changed line is a comment; verified by filtering the diff. Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

Chemaclass · 2026-08-02T09:06:46Z

🤔 Background

A # shellcheck disable= for a warning that no longer fires is dead weight — and misleading, since it tells the next reader the line below is doing something ShellCheck objects to when it is not. This project has found stale ones twice before (#928, #947), both times by accident rather than by looking.

💡 Changes

Removed 11 dead directives across 9 files: SC2153, SC2034 ×3, SC2154 ×3, SC2254, SC2030, SC2031, SC2295
Kept every SC1091 / SC2155 / SC2016 disable — those don't fire only because both gates exclude those codes globally, which is not the same as being obsolete; they still work for anyone running plain shellcheck
Explanatory comments above or beside a removed directive are kept where they still say something true

🔍 How they were found

Mechanically, not by eye: strip every disable from a file, re-run ShellCheck, see which declared codes come back. Two details decided correct-vs-wrong:

In place, not in /tmp..shellcheckrc sets source-path=src + external-sources=true, so copying a file elsewhere breaks source resolution and changes what SC2154/SC2034 report — the exact codes under test.
Both gates.make sa runs -xC (follows source); CI's ludeeus/action-shellcheck does not. They disagree on cross-file symbols, so a directive only counts as dead if it fires under neither. That didn't change the final list, but assuming it would have been luck.

✅ Verification

Every changed line is a comment (diff filtered to confirm). make sa · make lint · CI-mode ShellCheck (no -x, exact CI opts, clean) · bash build.sh bin -v → ✅ Build verified ✅.

Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

A directive for a warning that no longer fires is dead weight, and worse than that it is misleading -- it tells the next reader that the line below is doing something ShellCheck objects to, when it is not. This project has found stale ones twice before, in #928 and #947, both times by accident. Detected mechanically rather than by eye: strip every disable from a file, run ShellCheck, and see which declared codes actually come back. Two details made the difference between a correct list and a wrong one. The scan has to run in place. `.shellcheckrc` sets `source-path=src` with `external-sources=true`, so copying a file to a temp dir breaks source resolution and changes what SC2154 and SC2034 report -- exactly the codes under test. And it has to run under both gates. `make sa` uses `-xC` (follows `source`), CI's ludeeus/action-shellcheck does not, and the two see different things for cross-file symbols. A directive only counts as dead if it fires under neither. That distinction turned out not to change the list here, but assuming it would have been luck rather than method. Kept: the disables for SC1091, SC2155 and SC2016. Those never fire because both gates already exclude those codes globally, which is not the same as the warning being obsolete -- they still do their job for anyone running plain `shellcheck`. Removed (9 files, 11 lines): SC2153, SC2034 x3, SC2154 x3, SC2254, SC2030, SC2031, SC2295. Explanatory comments that sat above or beside a removed directive are kept where they still say something true. Every changed line is a comment; verified by filtering the diff. Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

Chemaclass · 2026-08-02T09:06:46Z

🤔 Background

A # shellcheck disable= for a warning that no longer fires is dead weight — and misleading, since it tells the next reader the line below is doing something ShellCheck objects to when it is not. This project has found stale ones twice before (#928, #947), both times by accident rather than by looking.

💡 Changes

Removed 11 dead directives across 9 files: SC2153, SC2034 ×3, SC2154 ×3, SC2254, SC2030, SC2031, SC2295
Kept every SC1091 / SC2155 / SC2016 disable — those don't fire only because both gates exclude those codes globally, which is not the same as being obsolete; they still work for anyone running plain shellcheck
Explanatory comments above or beside a removed directive are kept where they still say something true

🔍 How they were found

Mechanically, not by eye: strip every disable from a file, re-run ShellCheck, see which declared codes come back. Two details decided correct-vs-wrong:

In place, not in /tmp..shellcheckrc sets source-path=src + external-sources=true, so copying a file elsewhere breaks source resolution and changes what SC2154/SC2034 report — the exact codes under test.
Both gates.make sa runs -xC (follows source); CI's ludeeus/action-shellcheck does not. They disagree on cross-file symbols, so a directive only counts as dead if it fires under neither. That didn't change the final list, but assuming it would have been luck.

✅ Verification

Every changed line is a comment (diff filtered to confirm). make sa · make lint · CI-mode ShellCheck (no -x, exact CI opts, clean) · bash build.sh bin -v → ✅ Build verified ✅.

Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

A directive for a warning that no longer fires is dead weight, and worse than that it is misleading -- it tells the next reader that the line below is doing something ShellCheck objects to, when it is not. This project has found stale ones twice before, in #928 and #947, both times by accident. Detected mechanically rather than by eye: strip every disable from a file, run ShellCheck, and see which declared codes actually come back. Two details made the difference between a correct list and a wrong one. The scan has to run in place. `.shellcheckrc` sets `source-path=src` with `external-sources=true`, so copying a file to a temp dir breaks source resolution and changes what SC2154 and SC2034 report -- exactly the codes under test. And it has to run under both gates. `make sa` uses `-xC` (follows `source`), CI's ludeeus/action-shellcheck does not, and the two see different things for cross-file symbols. A directive only counts as dead if it fires under neither. That distinction turned out not to change the list here, but assuming it would have been luck rather than method. Kept: the disables for SC1091, SC2155 and SC2016. Those never fire because both gates already exclude those codes globally, which is not the same as the warning being obsolete -- they still do their job for anyone running plain `shellcheck`. Removed (9 files, 11 lines): SC2153, SC2034 x3, SC2154 x3, SC2254, SC2030, SC2031, SC2295. Explanatory comments that sat above or beside a removed directive are kept where they still say something true. Every changed line is a comment; verified by filtering the diff. Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

Chemaclass · 2026-08-02T09:06:46Z

🤔 Background

A # shellcheck disable= for a warning that no longer fires is dead weight — and misleading, since it tells the next reader the line below is doing something ShellCheck objects to when it is not. This project has found stale ones twice before (#928, #947), both times by accident rather than by looking.

💡 Changes

Removed 11 dead directives across 9 files: SC2153, SC2034 ×3, SC2154 ×3, SC2254, SC2030, SC2031, SC2295
Kept every SC1091 / SC2155 / SC2016 disable — those don't fire only because both gates exclude those codes globally, which is not the same as being obsolete; they still work for anyone running plain shellcheck
Explanatory comments above or beside a removed directive are kept where they still say something true

🔍 How they were found

Mechanically, not by eye: strip every disable from a file, re-run ShellCheck, see which declared codes come back. Two details decided correct-vs-wrong:

In place, not in /tmp..shellcheckrc sets source-path=src + external-sources=true, so copying a file elsewhere breaks source resolution and changes what SC2154/SC2034 report — the exact codes under test.
Both gates.make sa runs -xC (follows source); CI's ludeeus/action-shellcheck does not. They disagree on cross-file symbols, so a directive only counts as dead if it fires under neither. That didn't change the final list, but assuming it would have been luck.

✅ Verification

Every changed line is a comment (diff filtered to confirm). make sa · make lint · CI-mode ShellCheck (no -x, exact CI opts, clean) · bash build.sh bin -v → ✅ Build verified ✅.

Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

A directive for a warning that no longer fires is dead weight, and worse than that it is misleading -- it tells the next reader that the line below is doing something ShellCheck objects to, when it is not. This project has found stale ones twice before, in #928 and #947, both times by accident. Detected mechanically rather than by eye: strip every disable from a file, run ShellCheck, and see which declared codes actually come back. Two details made the difference between a correct list and a wrong one. The scan has to run in place. `.shellcheckrc` sets `source-path=src` with `external-sources=true`, so copying a file to a temp dir breaks source resolution and changes what SC2154 and SC2034 report -- exactly the codes under test. And it has to run under both gates. `make sa` uses `-xC` (follows `source`), CI's ludeeus/action-shellcheck does not, and the two see different things for cross-file symbols. A directive only counts as dead if it fires under neither. That distinction turned out not to change the list here, but assuming it would have been luck rather than method. Kept: the disables for SC1091, SC2155 and SC2016. Those never fire because both gates already exclude those codes globally, which is not the same as the warning being obsolete -- they still do their job for anyone running plain `shellcheck`. Removed (9 files, 11 lines): SC2153, SC2034 x3, SC2154 x3, SC2254, SC2030, SC2031, SC2295. Explanatory comments that sat above or beside a removed directive are kept where they still say something true. Every changed line is a comment; verified by filtering the diff. Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

Chemaclass · 2026-08-02T09:06:46Z

🤔 Background

A # shellcheck disable= for a warning that no longer fires is dead weight — and misleading, since it tells the next reader the line below is doing something ShellCheck objects to when it is not. This project has found stale ones twice before (#928, #947), both times by accident rather than by looking.

💡 Changes

Removed 11 dead directives across 9 files: SC2153, SC2034 ×3, SC2154 ×3, SC2254, SC2030, SC2031, SC2295
Kept every SC1091 / SC2155 / SC2016 disable — those don't fire only because both gates exclude those codes globally, which is not the same as being obsolete; they still work for anyone running plain shellcheck
Explanatory comments above or beside a removed directive are kept where they still say something true

🔍 How they were found

Mechanically, not by eye: strip every disable from a file, re-run ShellCheck, see which declared codes come back. Two details decided correct-vs-wrong:

In place, not in /tmp..shellcheckrc sets source-path=src + external-sources=true, so copying a file elsewhere breaks source resolution and changes what SC2154/SC2034 report — the exact codes under test.
Both gates.make sa runs -xC (follows source); CI's ludeeus/action-shellcheck does not. They disagree on cross-file symbols, so a directive only counts as dead if it fires under neither. That didn't change the final list, but assuming it would have been luck.

✅ Verification

Every changed line is a comment (diff filtered to confirm). make sa · make lint · CI-mode ShellCheck (no -x, exact CI opts, clean) · bash build.sh bin -v → ✅ Build verified ✅.

Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

A directive for a warning that no longer fires is dead weight, and worse than that it is misleading -- it tells the next reader that the line below is doing something ShellCheck objects to, when it is not. This project has found stale ones twice before, in #928 and #947, both times by accident. Detected mechanically rather than by eye: strip every disable from a file, run ShellCheck, and see which declared codes actually come back. Two details made the difference between a correct list and a wrong one. The scan has to run in place. `.shellcheckrc` sets `source-path=src` with `external-sources=true`, so copying a file to a temp dir breaks source resolution and changes what SC2154 and SC2034 report -- exactly the codes under test. And it has to run under both gates. `make sa` uses `-xC` (follows `source`), CI's ludeeus/action-shellcheck does not, and the two see different things for cross-file symbols. A directive only counts as dead if it fires under neither. That distinction turned out not to change the list here, but assuming it would have been luck rather than method. Kept: the disables for SC1091, SC2155 and SC2016. Those never fire because both gates already exclude those codes globally, which is not the same as the warning being obsolete -- they still do their job for anyone running plain `shellcheck`. Removed (9 files, 11 lines): SC2153, SC2034 x3, SC2154 x3, SC2254, SC2030, SC2031, SC2295. Explanatory comments that sat above or beside a removed directive are kept where they still say something true. Every changed line is a comment; verified by filtering the diff. Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

Chemaclass · 2026-08-02T09:06:46Z

🤔 Background

A # shellcheck disable= for a warning that no longer fires is dead weight — and misleading, since it tells the next reader the line below is doing something ShellCheck objects to when it is not. This project has found stale ones twice before (#928, #947), both times by accident rather than by looking.

💡 Changes

Removed 11 dead directives across 9 files: SC2153, SC2034 ×3, SC2154 ×3, SC2254, SC2030, SC2031, SC2295
Kept every SC1091 / SC2155 / SC2016 disable — those don't fire only because both gates exclude those codes globally, which is not the same as being obsolete; they still work for anyone running plain shellcheck
Explanatory comments above or beside a removed directive are kept where they still say something true

🔍 How they were found

Mechanically, not by eye: strip every disable from a file, re-run ShellCheck, see which declared codes come back. Two details decided correct-vs-wrong:

In place, not in /tmp..shellcheckrc sets source-path=src + external-sources=true, so copying a file elsewhere breaks source resolution and changes what SC2154/SC2034 report — the exact codes under test.
Both gates.make sa runs -xC (follows source); CI's ludeeus/action-shellcheck does not. They disagree on cross-file symbols, so a directive only counts as dead if it fires under neither. That didn't change the final list, but assuming it would have been luck.

✅ Verification

Every changed line is a comment (diff filtered to confirm). make sa · make lint · CI-mode ShellCheck (no -x, exact CI opts, clean) · bash build.sh bin -v → ✅ Build verified ✅.

Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

A directive for a warning that no longer fires is dead weight, and worse than that it is misleading -- it tells the next reader that the line below is doing something ShellCheck objects to, when it is not. This project has found stale ones twice before, in #928 and #947, both times by accident. Detected mechanically rather than by eye: strip every disable from a file, run ShellCheck, and see which declared codes actually come back. Two details made the difference between a correct list and a wrong one. The scan has to run in place. `.shellcheckrc` sets `source-path=src` with `external-sources=true`, so copying a file to a temp dir breaks source resolution and changes what SC2154 and SC2034 report -- exactly the codes under test. And it has to run under both gates. `make sa` uses `-xC` (follows `source`), CI's ludeeus/action-shellcheck does not, and the two see different things for cross-file symbols. A directive only counts as dead if it fires under neither. That distinction turned out not to change the list here, but assuming it would have been luck rather than method. Kept: the disables for SC1091, SC2155 and SC2016. Those never fire because both gates already exclude those codes globally, which is not the same as the warning being obsolete -- they still do their job for anyone running plain `shellcheck`. Removed (9 files, 11 lines): SC2153, SC2034 x3, SC2154 x3, SC2254, SC2030, SC2031, SC2295. Explanatory comments that sat above or beside a removed directive are kept where they still say something true. Every changed line is a comment; verified by filtering the diff. Suites match baseline exactly: 1630 sequential, 1589 parallel/simple/strict.

Chemaclass added the refactoring Refactoring or cleaning related label Aug 2, 2026

Chemaclass self-assigned this Aug 2, 2026

Chemaclass merged commit fa9470c into mainAug 2, 2026
37 checks passed