refactor(system): group the capability probes into a src/system/ module by Chemaclass · Pull Request #950 · TypedDevs/bashunit

Chemaclass · 2026-08-01T13:17:55Z

🤔 Background

Related #949 — first group.

src/ held 11 modules and 15 loose files. Individually each is small and well named; as a listing they read as leftovers and the root stopped conveying structure.

💡 Changes

src/check_os.sh -> src/system/check_os.sh
src/dependencies.sh -> src/system/dependencies.sh
src/io.sh -> src/system/io.sh

The bottom layer — what this machine is and what it has. Nothing in it touches config, test state, console or the runner. All three recorded as renames with a zero-line diff.

⚠️ Correction to the issue's proposed grouping

#949 put clock.sh here too. That creates a cross-module cycle:

system/clock.sh -> util/math.sh -> system/dependencies.sh

Harmless at load time — both crossings are inside function bodies, and clock.sh's only file-scope statements are two variable initialisers — but a module cycle isn't worth shipping when one different cut removes it. clock.sh goes to src/util/ with math.sh instead: both compute something, both probe via dependencies, so the edge runs util → system in one direction only. Mapping updated on the issue.

📌 Two callers reach these by path, not through the entrypoint

build.sh:4 sources check_os.sh and calls check_os::initbefore it sets BASHUNIT_ROOT_DIR — so it points at the file, not the module index.
tests/unit/check_os_test.sh:88 sources it relative to the test file.

Both would have failed silently or loudly on merge; grepped for before committing, per the checklist #938 and #946 earned.

✅ Verification

Artifact code content identical, bash build.sh bin -v → ✅ Build verified ✅.

Green: sequential (1604 passed, 0 failed) · --parallel --simple --strict · make sa · make lint · CI-mode ShellCheck · git check-ignore clean.

First group under #949. src/ held eleven modules and fifteen loose files; individually each was small and well named, but as a listing they read as leftovers and the root stopped conveying structure. src/check_os.sh -> src/system/check_os.sh src/dependencies.sh -> src/system/dependencies.sh src/io.sh -> src/system/io.sh The bottom layer: what this machine is and what it has. Nothing in it touches config, test state, console or the runner. The issue proposed putting clock.sh here too. That would have created a cross-module cycle -- system/clock.sh -> util/math.sh -> system/dependencies.sh. Harmless at load time, since both crossings are inside function bodies and clock.sh's only file-scope statements are two variable initialisers, but not worth shipping. clock.sh goes to src/util/ with math.sh instead: both compute something and both probe via dependencies, so the edge runs util -> system in one direction only. Two callers reach these files by path rather than through the entrypoint: build.sh sources check_os.sh and calls check_os::init at line 4, before it sets BASHUNIT_ROOT_DIR, so it points at the file rather than the module index; and tests/unit/check_os_test.sh sources it relative to the test file. A relocation: git records all three as renames with a zero-line diff, and the built artifact's code content is identical. Related #949

Chemaclass · 2026-08-01T13:17:55Z

🤔 Background

Related #949 — first group.

src/ held 11 modules and 15 loose files. Individually each is small and well named; as a listing they read as leftovers and the root stopped conveying structure.

💡 Changes

src/check_os.sh -> src/system/check_os.sh
src/dependencies.sh -> src/system/dependencies.sh
src/io.sh -> src/system/io.sh

The bottom layer — what this machine is and what it has. Nothing in it touches config, test state, console or the runner. All three recorded as renames with a zero-line diff.

⚠️ Correction to the issue's proposed grouping

#949 put clock.sh here too. That creates a cross-module cycle:

system/clock.sh -> util/math.sh -> system/dependencies.sh

Harmless at load time — both crossings are inside function bodies, and clock.sh's only file-scope statements are two variable initialisers — but a module cycle isn't worth shipping when one different cut removes it. clock.sh goes to src/util/ with math.sh instead: both compute something, both probe via dependencies, so the edge runs util → system in one direction only. Mapping updated on the issue.

📌 Two callers reach these by path, not through the entrypoint

build.sh:4 sources check_os.sh and calls check_os::initbefore it sets BASHUNIT_ROOT_DIR — so it points at the file, not the module index.
tests/unit/check_os_test.sh:88 sources it relative to the test file.

Both would have failed silently or loudly on merge; grepped for before committing, per the checklist #938 and #946 earned.

✅ Verification

Artifact code content identical, bash build.sh bin -v → ✅ Build verified ✅.

Green: sequential (1604 passed, 0 failed) · --parallel --simple --strict · make sa · make lint · CI-mode ShellCheck · git check-ignore clean.

First group under #949. src/ held eleven modules and fifteen loose files; individually each was small and well named, but as a listing they read as leftovers and the root stopped conveying structure. src/check_os.sh -> src/system/check_os.sh src/dependencies.sh -> src/system/dependencies.sh src/io.sh -> src/system/io.sh The bottom layer: what this machine is and what it has. Nothing in it touches config, test state, console or the runner. The issue proposed putting clock.sh here too. That would have created a cross-module cycle -- system/clock.sh -> util/math.sh -> system/dependencies.sh. Harmless at load time, since both crossings are inside function bodies and clock.sh's only file-scope statements are two variable initialisers, but not worth shipping. clock.sh goes to src/util/ with math.sh instead: both compute something and both probe via dependencies, so the edge runs util -> system in one direction only. Two callers reach these files by path rather than through the entrypoint: build.sh sources check_os.sh and calls check_os::init at line 4, before it sets BASHUNIT_ROOT_DIR, so it points at the file rather than the module index; and tests/unit/check_os_test.sh sources it relative to the test file. A relocation: git records all three as renames with a zero-line diff, and the built artifact's code content is identical. Related #949

Chemaclass · 2026-08-01T13:17:55Z

🤔 Background

Related #949 — first group.

src/ held 11 modules and 15 loose files. Individually each is small and well named; as a listing they read as leftovers and the root stopped conveying structure.

💡 Changes

src/check_os.sh -> src/system/check_os.sh
src/dependencies.sh -> src/system/dependencies.sh
src/io.sh -> src/system/io.sh

The bottom layer — what this machine is and what it has. Nothing in it touches config, test state, console or the runner. All three recorded as renames with a zero-line diff.

⚠️ Correction to the issue's proposed grouping

#949 put clock.sh here too. That creates a cross-module cycle:

system/clock.sh -> util/math.sh -> system/dependencies.sh

Harmless at load time — both crossings are inside function bodies, and clock.sh's only file-scope statements are two variable initialisers — but a module cycle isn't worth shipping when one different cut removes it. clock.sh goes to src/util/ with math.sh instead: both compute something, both probe via dependencies, so the edge runs util → system in one direction only. Mapping updated on the issue.

📌 Two callers reach these by path, not through the entrypoint

build.sh:4 sources check_os.sh and calls check_os::initbefore it sets BASHUNIT_ROOT_DIR — so it points at the file, not the module index.
tests/unit/check_os_test.sh:88 sources it relative to the test file.

Both would have failed silently or loudly on merge; grepped for before committing, per the checklist #938 and #946 earned.

✅ Verification

Artifact code content identical, bash build.sh bin -v → ✅ Build verified ✅.

Green: sequential (1604 passed, 0 failed) · --parallel --simple --strict · make sa · make lint · CI-mode ShellCheck · git check-ignore clean.

First group under #949. src/ held eleven modules and fifteen loose files; individually each was small and well named, but as a listing they read as leftovers and the root stopped conveying structure. src/check_os.sh -> src/system/check_os.sh src/dependencies.sh -> src/system/dependencies.sh src/io.sh -> src/system/io.sh The bottom layer: what this machine is and what it has. Nothing in it touches config, test state, console or the runner. The issue proposed putting clock.sh here too. That would have created a cross-module cycle -- system/clock.sh -> util/math.sh -> system/dependencies.sh. Harmless at load time, since both crossings are inside function bodies and clock.sh's only file-scope statements are two variable initialisers, but not worth shipping. clock.sh goes to src/util/ with math.sh instead: both compute something and both probe via dependencies, so the edge runs util -> system in one direction only. Two callers reach these files by path rather than through the entrypoint: build.sh sources check_os.sh and calls check_os::init at line 4, before it sets BASHUNIT_ROOT_DIR, so it points at the file rather than the module index; and tests/unit/check_os_test.sh sources it relative to the test file. A relocation: git records all three as renames with a zero-line diff, and the built artifact's code content is identical. Related #949

Chemaclass · 2026-08-01T13:17:55Z

🤔 Background

Related #949 — first group.

src/ held 11 modules and 15 loose files. Individually each is small and well named; as a listing they read as leftovers and the root stopped conveying structure.

💡 Changes

src/check_os.sh -> src/system/check_os.sh
src/dependencies.sh -> src/system/dependencies.sh
src/io.sh -> src/system/io.sh

The bottom layer — what this machine is and what it has. Nothing in it touches config, test state, console or the runner. All three recorded as renames with a zero-line diff.

⚠️ Correction to the issue's proposed grouping

#949 put clock.sh here too. That creates a cross-module cycle:

system/clock.sh -> util/math.sh -> system/dependencies.sh

Harmless at load time — both crossings are inside function bodies, and clock.sh's only file-scope statements are two variable initialisers — but a module cycle isn't worth shipping when one different cut removes it. clock.sh goes to src/util/ with math.sh instead: both compute something, both probe via dependencies, so the edge runs util → system in one direction only. Mapping updated on the issue.

📌 Two callers reach these by path, not through the entrypoint

build.sh:4 sources check_os.sh and calls check_os::initbefore it sets BASHUNIT_ROOT_DIR — so it points at the file, not the module index.
tests/unit/check_os_test.sh:88 sources it relative to the test file.

Both would have failed silently or loudly on merge; grepped for before committing, per the checklist #938 and #946 earned.

✅ Verification

Artifact code content identical, bash build.sh bin -v → ✅ Build verified ✅.

Green: sequential (1604 passed, 0 failed) · --parallel --simple --strict · make sa · make lint · CI-mode ShellCheck · git check-ignore clean.

First group under #949. src/ held eleven modules and fifteen loose files; individually each was small and well named, but as a listing they read as leftovers and the root stopped conveying structure. src/check_os.sh -> src/system/check_os.sh src/dependencies.sh -> src/system/dependencies.sh src/io.sh -> src/system/io.sh The bottom layer: what this machine is and what it has. Nothing in it touches config, test state, console or the runner. The issue proposed putting clock.sh here too. That would have created a cross-module cycle -- system/clock.sh -> util/math.sh -> system/dependencies.sh. Harmless at load time, since both crossings are inside function bodies and clock.sh's only file-scope statements are two variable initialisers, but not worth shipping. clock.sh goes to src/util/ with math.sh instead: both compute something and both probe via dependencies, so the edge runs util -> system in one direction only. Two callers reach these files by path rather than through the entrypoint: build.sh sources check_os.sh and calls check_os::init at line 4, before it sets BASHUNIT_ROOT_DIR, so it points at the file rather than the module index; and tests/unit/check_os_test.sh sources it relative to the test file. A relocation: git records all three as renames with a zero-line diff, and the built artifact's code content is identical. Related #949

Chemaclass · 2026-08-01T13:17:55Z

🤔 Background

Related #949 — first group.

src/ held 11 modules and 15 loose files. Individually each is small and well named; as a listing they read as leftovers and the root stopped conveying structure.

💡 Changes

src/check_os.sh -> src/system/check_os.sh
src/dependencies.sh -> src/system/dependencies.sh
src/io.sh -> src/system/io.sh

The bottom layer — what this machine is and what it has. Nothing in it touches config, test state, console or the runner. All three recorded as renames with a zero-line diff.

⚠️ Correction to the issue's proposed grouping

#949 put clock.sh here too. That creates a cross-module cycle:

system/clock.sh -> util/math.sh -> system/dependencies.sh

Harmless at load time — both crossings are inside function bodies, and clock.sh's only file-scope statements are two variable initialisers — but a module cycle isn't worth shipping when one different cut removes it. clock.sh goes to src/util/ with math.sh instead: both compute something, both probe via dependencies, so the edge runs util → system in one direction only. Mapping updated on the issue.

📌 Two callers reach these by path, not through the entrypoint

build.sh:4 sources check_os.sh and calls check_os::initbefore it sets BASHUNIT_ROOT_DIR — so it points at the file, not the module index.
tests/unit/check_os_test.sh:88 sources it relative to the test file.

Both would have failed silently or loudly on merge; grepped for before committing, per the checklist #938 and #946 earned.

✅ Verification

Artifact code content identical, bash build.sh bin -v → ✅ Build verified ✅.

Green: sequential (1604 passed, 0 failed) · --parallel --simple --strict · make sa · make lint · CI-mode ShellCheck · git check-ignore clean.

First group under #949. src/ held eleven modules and fifteen loose files; individually each was small and well named, but as a listing they read as leftovers and the root stopped conveying structure. src/check_os.sh -> src/system/check_os.sh src/dependencies.sh -> src/system/dependencies.sh src/io.sh -> src/system/io.sh The bottom layer: what this machine is and what it has. Nothing in it touches config, test state, console or the runner. The issue proposed putting clock.sh here too. That would have created a cross-module cycle -- system/clock.sh -> util/math.sh -> system/dependencies.sh. Harmless at load time, since both crossings are inside function bodies and clock.sh's only file-scope statements are two variable initialisers, but not worth shipping. clock.sh goes to src/util/ with math.sh instead: both compute something and both probe via dependencies, so the edge runs util -> system in one direction only. Two callers reach these files by path rather than through the entrypoint: build.sh sources check_os.sh and calls check_os::init at line 4, before it sets BASHUNIT_ROOT_DIR, so it points at the file rather than the module index; and tests/unit/check_os_test.sh sources it relative to the test file. A relocation: git records all three as renames with a zero-line diff, and the built artifact's code content is identical. Related #949

Chemaclass · 2026-08-01T13:17:55Z

🤔 Background

Related #949 — first group.

src/ held 11 modules and 15 loose files. Individually each is small and well named; as a listing they read as leftovers and the root stopped conveying structure.

💡 Changes

src/check_os.sh -> src/system/check_os.sh
src/dependencies.sh -> src/system/dependencies.sh
src/io.sh -> src/system/io.sh

The bottom layer — what this machine is and what it has. Nothing in it touches config, test state, console or the runner. All three recorded as renames with a zero-line diff.

⚠️ Correction to the issue's proposed grouping

#949 put clock.sh here too. That creates a cross-module cycle:

system/clock.sh -> util/math.sh -> system/dependencies.sh

Harmless at load time — both crossings are inside function bodies, and clock.sh's only file-scope statements are two variable initialisers — but a module cycle isn't worth shipping when one different cut removes it. clock.sh goes to src/util/ with math.sh instead: both compute something, both probe via dependencies, so the edge runs util → system in one direction only. Mapping updated on the issue.

📌 Two callers reach these by path, not through the entrypoint

build.sh:4 sources check_os.sh and calls check_os::initbefore it sets BASHUNIT_ROOT_DIR — so it points at the file, not the module index.
tests/unit/check_os_test.sh:88 sources it relative to the test file.

Both would have failed silently or loudly on merge; grepped for before committing, per the checklist #938 and #946 earned.

✅ Verification

Artifact code content identical, bash build.sh bin -v → ✅ Build verified ✅.

Green: sequential (1604 passed, 0 failed) · --parallel --simple --strict · make sa · make lint · CI-mode ShellCheck · git check-ignore clean.

First group under #949. src/ held eleven modules and fifteen loose files; individually each was small and well named, but as a listing they read as leftovers and the root stopped conveying structure. src/check_os.sh -> src/system/check_os.sh src/dependencies.sh -> src/system/dependencies.sh src/io.sh -> src/system/io.sh The bottom layer: what this machine is and what it has. Nothing in it touches config, test state, console or the runner. The issue proposed putting clock.sh here too. That would have created a cross-module cycle -- system/clock.sh -> util/math.sh -> system/dependencies.sh. Harmless at load time, since both crossings are inside function bodies and clock.sh's only file-scope statements are two variable initialisers, but not worth shipping. clock.sh goes to src/util/ with math.sh instead: both compute something and both probe via dependencies, so the edge runs util -> system in one direction only. Two callers reach these files by path rather than through the entrypoint: build.sh sources check_os.sh and calls check_os::init at line 4, before it sets BASHUNIT_ROOT_DIR, so it points at the file rather than the module index; and tests/unit/check_os_test.sh sources it relative to the test file. A relocation: git records all three as renames with a zero-line diff, and the built artifact's code content is identical. Related #949

Chemaclass · 2026-08-01T13:17:55Z

🤔 Background

Related #949 — first group.

src/ held 11 modules and 15 loose files. Individually each is small and well named; as a listing they read as leftovers and the root stopped conveying structure.

💡 Changes

src/check_os.sh -> src/system/check_os.sh
src/dependencies.sh -> src/system/dependencies.sh
src/io.sh -> src/system/io.sh

The bottom layer — what this machine is and what it has. Nothing in it touches config, test state, console or the runner. All three recorded as renames with a zero-line diff.

⚠️ Correction to the issue's proposed grouping

#949 put clock.sh here too. That creates a cross-module cycle:

system/clock.sh -> util/math.sh -> system/dependencies.sh

Harmless at load time — both crossings are inside function bodies, and clock.sh's only file-scope statements are two variable initialisers — but a module cycle isn't worth shipping when one different cut removes it. clock.sh goes to src/util/ with math.sh instead: both compute something, both probe via dependencies, so the edge runs util → system in one direction only. Mapping updated on the issue.

📌 Two callers reach these by path, not through the entrypoint

build.sh:4 sources check_os.sh and calls check_os::initbefore it sets BASHUNIT_ROOT_DIR — so it points at the file, not the module index.
tests/unit/check_os_test.sh:88 sources it relative to the test file.

Both would have failed silently or loudly on merge; grepped for before committing, per the checklist #938 and #946 earned.

✅ Verification

Artifact code content identical, bash build.sh bin -v → ✅ Build verified ✅.

Green: sequential (1604 passed, 0 failed) · --parallel --simple --strict · make sa · make lint · CI-mode ShellCheck · git check-ignore clean.

First group under #949. src/ held eleven modules and fifteen loose files; individually each was small and well named, but as a listing they read as leftovers and the root stopped conveying structure. src/check_os.sh -> src/system/check_os.sh src/dependencies.sh -> src/system/dependencies.sh src/io.sh -> src/system/io.sh The bottom layer: what this machine is and what it has. Nothing in it touches config, test state, console or the runner. The issue proposed putting clock.sh here too. That would have created a cross-module cycle -- system/clock.sh -> util/math.sh -> system/dependencies.sh. Harmless at load time, since both crossings are inside function bodies and clock.sh's only file-scope statements are two variable initialisers, but not worth shipping. clock.sh goes to src/util/ with math.sh instead: both compute something and both probe via dependencies, so the edge runs util -> system in one direction only. Two callers reach these files by path rather than through the entrypoint: build.sh sources check_os.sh and calls check_os::init at line 4, before it sets BASHUNIT_ROOT_DIR, so it points at the file rather than the module index; and tests/unit/check_os_test.sh sources it relative to the test file. A relocation: git records all three as renames with a zero-line diff, and the built artifact's code content is identical. Related #949

Chemaclass · 2026-08-01T13:17:55Z

🤔 Background

Related #949 — first group.

src/ held 11 modules and 15 loose files. Individually each is small and well named; as a listing they read as leftovers and the root stopped conveying structure.

💡 Changes

src/check_os.sh -> src/system/check_os.sh
src/dependencies.sh -> src/system/dependencies.sh
src/io.sh -> src/system/io.sh

The bottom layer — what this machine is and what it has. Nothing in it touches config, test state, console or the runner. All three recorded as renames with a zero-line diff.

⚠️ Correction to the issue's proposed grouping

#949 put clock.sh here too. That creates a cross-module cycle:

system/clock.sh -> util/math.sh -> system/dependencies.sh

Harmless at load time — both crossings are inside function bodies, and clock.sh's only file-scope statements are two variable initialisers — but a module cycle isn't worth shipping when one different cut removes it. clock.sh goes to src/util/ with math.sh instead: both compute something, both probe via dependencies, so the edge runs util → system in one direction only. Mapping updated on the issue.

📌 Two callers reach these by path, not through the entrypoint

build.sh:4 sources check_os.sh and calls check_os::initbefore it sets BASHUNIT_ROOT_DIR — so it points at the file, not the module index.
tests/unit/check_os_test.sh:88 sources it relative to the test file.

Both would have failed silently or loudly on merge; grepped for before committing, per the checklist #938 and #946 earned.

✅ Verification

Artifact code content identical, bash build.sh bin -v → ✅ Build verified ✅.

Green: sequential (1604 passed, 0 failed) · --parallel --simple --strict · make sa · make lint · CI-mode ShellCheck · git check-ignore clean.

First group under #949. src/ held eleven modules and fifteen loose files; individually each was small and well named, but as a listing they read as leftovers and the root stopped conveying structure. src/check_os.sh -> src/system/check_os.sh src/dependencies.sh -> src/system/dependencies.sh src/io.sh -> src/system/io.sh The bottom layer: what this machine is and what it has. Nothing in it touches config, test state, console or the runner. The issue proposed putting clock.sh here too. That would have created a cross-module cycle -- system/clock.sh -> util/math.sh -> system/dependencies.sh. Harmless at load time, since both crossings are inside function bodies and clock.sh's only file-scope statements are two variable initialisers, but not worth shipping. clock.sh goes to src/util/ with math.sh instead: both compute something and both probe via dependencies, so the edge runs util -> system in one direction only. Two callers reach these files by path rather than through the entrypoint: build.sh sources check_os.sh and calls check_os::init at line 4, before it sets BASHUNIT_ROOT_DIR, so it points at the file rather than the module index; and tests/unit/check_os_test.sh sources it relative to the test file. A relocation: git records all three as renames with a zero-line diff, and the built artifact's code content is identical. Related #949

Chemaclass added the refactoring Refactoring or cleaning related label Aug 1, 2026

Chemaclass self-assigned this Aug 1, 2026