Repository navigation
test(lab): record the lab acceptance and test every lab-only command - #6
Merged
Merged
Conversation
Pull request #5 merged as 2ebb3a5, and the release run published v0.3.0-preview0002 to GitHub and the PowerShell Gallery with the wiki updated. The record keeps only what is still open: live lab acceptance before the next stable release, the changelog token permission, and the recorded follow-ups. Both files end well under their line budget. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The development machine has no lab VMs, so handoff 07 runs on the Hyper-V host that holds WindowsAccessControlLab. The handoff folder travels there with a git bundle of main, its tags, and this record branch, and the record now names the development machine explicitly so it reads correctly on the lab host. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Live domain-lab acceptance of main 2ebb3a5, published as v0.3.0-preview0002, on the Hyper-V lab host. The release-equivalent build from GitVersion 5.12.0 matches the Gallery package byte-for-byte. Four lab passes ran 97 cases each with eight ready cleanup entries, the isolated Desktop DSC engine gate ran 5 of 5, and both local gates passed with the new lab coverage merged. The lab-runner ownership refusals were proven against decoys, and a byte and ACL inventory proves the restoration. Records the evidence in docs/lab-acceptance-2026-10-07.md, indexes it, and updates the Memory Bank with the lab state, open work, and lessons. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Four live cases reach the paths that only unit tests covered. A disposable Task Scheduler folder and task, whose protected DACLs lack Local System, accept an unrelated ACE, refuse a Local System deny without writing, and get Local System back. Around an SMB share DACL write, a raw native write shows that the description is cleared, and the command restores and reports it; a native watcher's concurrent edit is kept with a warning; and an injected failure of the setter's own description read stops the command before anything is written. Each case fails against the builds of 96d6671 and 5531824 for the reason its change names, and passes on the candidate. The specifications, the traceability table, the checklist, and the lab guide name the new evidence. The SMB usage page and specification 0009 now describe both windows in which a concurrent description edit can still be replaced. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The 2026-10-07 acceptance record now holds the follow-up: the four live cases, their red results against the parent builds, all four passes repeated with them, the final lab coverage and local gates, and three observations. The Memory Bank records the lab state, the open work, and the lessons from reaching each guard live. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…h seam Get-WindowsADRootDse and Get-WindowsADEffectiveAccessRecord sent their LDAP requests straight to the connection, so no unit test could reach their response checks and both files were declared domain-lab-only for coverage. Both now go through Send-WindowsADSearchRequest, like every other directory read. Unit tests replace the request at that seam and prove that a RootDSE or effective-access response without exactly one entry is refused, that a RootDSE without a root domain naming context reports none, and that an absent sDRightsEffective reports zero. The two files leave the domain-lab-only list, so the local coverage gate now asserts them. An object deleted between target resolution and the effective-access read now fails with an ItemNotFoundException naming it instead of a MethodInvocationException from the LDAP call. Specification 0018 and the changelog say so. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Eight live cases reach the SMB share and Active Directory paths that no test executed. On the member server, an SDDL without a DACL is refused before anything is written, -PassThru returns the stored descriptor, a rule copy that was not read from the share and a rule whose canonical target names another server are refused with the grant left in place, and a share DACL write without -Confirm:$false asks for confirmation and writes nothing when the caller cannot answer. That last case opens its own session from a runspace without a user interface, because a remote prompt goes to the host that opened the session, and it arms member coverage there under its own name. In the directory suite, a module-scope mock fails the inheritance-source or schema-name lookup after the real descriptor read, and the rules are still reported with a non-terminating error. The calls state the default Continue preference because the runner stops on every error. A disposable organizational unit deleted inside the effective-access request is reported as not found. Each case fails against a build with its guard removed and passes on the candidate. The checklist, the traceability table, and specification 0009 name the new evidence. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The Desktop coverage pass of 938bdff stopped at the confirmation case. A session opened from a runspace without a user interface starts with remote debugging off, so the coverage breakpoints could not be armed in it; the focused runs had no coverage armed and could not show this. The case now enables remote debugging on that session before arming coverage, and a focused Desktop run with member coverage armed records the confirmation's throttle line. Arming member coverage also leaves the session stopping on every error, because the harness sets the preference in a script block that the session runs at its top level. The case therefore states the default Continue preference under which the refusal is reported. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The 2026-10-08 record names the 17 commands that no test executed in the declared domain-lab-only files, how each now gets a test, the red evidence against a build with each guard removed, and the passes of bad1302: the DSC gate, four passes of 109 cases, and both local gates, with the domain-lab-only coverage at 100 percent. It also records why the first built pass of 938bdff stopped, the remote-prompt and error-preference behavior the cases rely on, and the cleanup. The 2026-10-07 record and the documentation index link to it, and the Memory Bank records the state, the open harness defect, and the lessons. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
Removing the older lab checkpoint with Remove-LabVMSnapshot also deleted the newer one that the maintainer had asked to keep, on ten of the thirteen VMs, because AutomatedLab removes a checkpoint with all its children. With the maintainer's agreement a new checkpoint, wac07-after-bad1302-9d9776ca, now covers all thirteen VMs, and the old checkpoints on the other three were merged away one at a time. The record, the Memory Bank lab state, and the lessons say so. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
…erage Enter-WindowsAccessControlMemberCoverage armed the member module through Invoke-Command -Session, which runs its script block at the top level of the persistent member session. The $ErrorActionPreference = 'Stop' it set there outlasted the call, so in the Desktop coverage pass the SMB share, certificate private-key, Task Scheduler, and foreign-principal suites stopped at their first non-terminating error, while every other pass continued. The arming body now runs in a child scope that receives its values as parameters; the global breakpoint list still outlives the call. A unit case runs the remote block in a separate runspace and asserts the armed count and the unchanged preference; it failed with 'Stop' against the previous harness. A live probe from F1ADC1 into an F1AFile1 member session showed 'Stop' and a call ended by a non-terminating error with the previous harness, and 'Continue' with the next statement reached after this change, in both editions, with all 8,504 locations armed. The SMB confirmation case keeps stating -ErrorAction Continue; its comment no longer cites the leak. The changelog and specification 0005 record the behavior. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
The 2026-10-08 acceptance record gains a Harness correction section for a572d3d: the unit and live red/green evidence, the candidate hashes with the unchanged root module, the built pass of 109 cases in both editions with line-identical lab coverage, both local gates (Core 1,938 at 91.26 percent and Desktop 1,893 at 90.53 percent asserted, domain-lab-only 100 percent), and the restoration. The observation and the retention paragraph no longer call the leak open work, and the docs index names the correction. The Memory Bank records the correction, drops the leak from the open work, states the snapshot-removal lesson as the checkpoint subtree, and moves the four oldest 2026-10-06 progress entries behind a git show pointer to stay within the line budget. Co-authored-by: AI Assistant <ai@example.com> Co-authored-by: Copilot <223556219+Copilot@users.noreply.github.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Pull Request (PR) description
This pull request lands the domain-lab acceptance of
v0.3.0-preview0002:its records, the tests it added, and the one module change it asked for. Each
part has its own commits, so any part can be reverted on its own.
Get-WindowsADRootDseandGet-WindowsADEffectiveAccessRecordnow read throughSend-WindowsADSearchRequest, the directory search function that everyother directory read already uses, so unit tests reach their four response
checks without a domain controller. Both files leave
CodeCoverage.DomainLabOnlySourcePathinbuild.yaml, so their commandscount toward the asserted coverage now (
4096d02).effective-access read of
Get-ADObjectCallerEffectiveAccessnow fails withan
ItemNotFoundExceptionnaming it, like the family's other reads, insteadof a
MethodInvocationExceptionfrom the LDAP call. Specification 0018states it in its output contract, and
CHANGELOG.mdrecords it underChanged(4096d02).a test. New live cases cover the Task Scheduler repair of a protected DACL
without Local System; the SMB share description that the native write
clears, that is edited concurrently, or that cannot be read before the
write; five SMB share guards, including the confirmation prompt in a session
without a user interface; two injected enrichment failures; and a directory
object deleted inside the effective-access read. Each case failed against a
build with its guard removed and passed against the candidate (
6ea9d7e,938bdff,bad1302).stopping on every error.
Enter-WindowsAccessControlMemberCoverageruns itsremote body in a child scope, so the
Stoppreference it sets ends with thecall (
a572d3d, changelogFixed).windows in which a share DACL write can still replace a concurrent
description edit (
6ea9d7e, changelogFixed). Specifications 0005, 0009,0010, and 0018 name the new evidence.
docs/lab-acceptance-2026-10-07.mdanddocs/lab-acceptance-2026-10-08.mdrecord the acceptance runs, and theMemory Bank records the publication of
v0.3.0-preview0002and theacceptance (
ce22f00,a18e52c,69dbfcd,6f6bd0b,c1c53cf,300e937,a95321c).No commit message on this branch matches a GitVersion bump pattern.
GitVersion 5.12.0, run on a simulated merge commit of this branch into
main, reports0.3.0-preview0003, so the merge publishes a preview.This Pull Request (PR) fixes the following issues
What was run, and what the result was
On
a95321c, on Windows with PowerShell 7.6.6 and Windows PowerShell5.1.26100, each step in its own process. No domain-lab coverage document was
present, so the coverage figures are the ones CI computes.
.\build.ps1 -Tasks build.\build.ps1 -Tasks testin PowerShell 7.\build.ps1 -Tasks testin Windows PowerShell 5.1Invoke-ScriptAnalyzer1.25.0 over the 9 changed PowerShell filesThe two skipped tests are the documented environment cases: no mounted volume
on the host, and no privilege to read the audit section.
The domain acceptance lab ran on 2026-10-07 and 2026-10-08:
docs/lab-acceptance-2026-10-07.md:mainat2ebb3a5,built as
0.3.0-preview0002and byte-identical to the Gallery package. TheDesktop DSC-engine gate passed 5 of 5, four full passes (built module and
installed package, Desktop and Core) passed 97 of 97 each, and both
lab-runner ownership refusals were proven against decoys. A follow-up added
four live cases, each red against the code before its change, and repeated
the four passes at 101 of 101.
docs/lab-acceptance-2026-10-08.md: candidatebad1302.The DSC-engine gate passed 5 of 5, and four passes of 109 cases each passed
with eight ready cleanup entries. With the fresh lab coverage merged, the
local gates passed 1,937 tests in PowerShell 7 at 91.26 % asserted coverage
and 1,892 in Windows PowerShell 5.1 at 90.53 %; domain-lab-only coverage is
100 % (154 of 154 commands). The harness correction
a572d3d, with themodule bytes unchanged, passed a built pass of 109 of 109 in both editions
with line-identical lab coverage, and the local gates passed 1,938 and 1,893
tests at the same coverage.
a95321cchanges only records.Task list
CHANGELOG.md. The entry says what changed for someone using the module and
why, and references the issue being resolved (if applicable).
specs/, everyrequirement identifier it adds is unique, and each one traces to
executable evidence in
specs/0005-verification-and-traceability.md.specs/decisions/andindexed in
specs/decisions/README.md.including a synopsis, a description, an example, and every parameter.
README.mdor underdocs/.fails without the fix.
measured against a real object.
.\build.ps1 -Tasks testpasses in both supported PowerShell editions,including the coverage threshold, with no assertion weakened to get there.
Invoke-ScriptAnalyzeris clean over everything that changed, and anysuppression carries a
Justification.