CI: Retry powershell commands on transient AccessViolationExceptions

[smalis-msft]

Problem

Petri-based VMM tests intermittently fail on Windows CI runners with powershell.exe crashing during session initialization:

exit code: 0xDEAD
System.AccessViolationException ... at EventLogLogProvider.LogProviderLifecycleEvent

The crash happens in InitialSessionState.Bind_LoadProviders before the user's command is dispatched, so retrying is idempotent.

Approach

Bake a tightly-scoped retry into PowerShellBuilder:

• support/powershell_builder gains PowerShellBuilder::output(). These run the command and retry up to twice on the exact AV signature (exit 0xDEAD + stderr containing both System.AccessViolationException and EventLogLogProvider). Other failures are returned immediately.
• petri::run_host_ps(builder) is the PowerShell entry point used by all hyperv/powershell.rs cmdlet wrappers. It pipes stdio + logs identically to run_host_cmd but goes through PowerShellBuilder::output(), so retry is automatic for every callsite.
• petri::run_host_cmd(Command) is unchanged for non-PowerShell tools (hvc.exe, vmgs).
• flowey_lib_hvlite::run_prep_steps likewise calls .output() on the builder.

The retry signature is precise enough that no false positives are expected, and the retried call is the same Command invocation — no partial side effects to worry about.

Alternatives considered

• pwsh.exe (PowerShell 7): blocked — runner images don't have it preinstalled.
• Disabling Windows PowerShell engine event logging via registry (EnableEventLogging=0): verified ineffective in CI run 25811224088 — LogProviderLifecycleEvent fires regardless of that switch.
• Installing PS7 on runners: rejected (don't want a new image dep).

[Copilot]

Pull request overview

Note

Copilot was unable to run its full agentic suite in this review.

Adds a Flowey CI step to disable Windows PowerShell engine event logging on Windows runners to mitigate transient powershell.exe startup crashes that cause test flakiness.

Changes:

• Introduces a new Flowey node to harden PowerShell event logging via registry settings.
• Wires the hardening node into the Windows test job’s pre-run dependencies.
• Adds explicit CI workflow steps to run the hardening node before starting test services.

Reviewed changes

Copilot reviewed 7 out of 7 changed files in this pull request and generated 6 comments.

Show a summary per file

File	Description
flowey/flowey_lib_hvlite/src/lib.rs	Exposes the new hardening module from the crate.
flowey/flowey_lib_hvlite/src/harden_powershell_event_log.rs	Implements the Windows runner hardening step (registry + event source check).
flowey/flowey_lib_hvlite/src/_jobs/consume_and_test_nextest_vmm_tests_archive.rs	Ensures hardening runs before tests on Windows.
ci-flowey/openvmm-pr.yaml	Adds pipeline steps to run the hardening node before starting test services.
.github/workflows/openvmm-pr.yaml	Adds GitHub Actions steps to run the hardening node in multiple Windows jobs.
.github/workflows/openvmm-pr-release.yaml	Adds the hardening step to Windows release PR workflow jobs.
.github/workflows/openvmm-ci.yaml	Adds the hardening step to Windows CI workflow jobs.

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated 1 comment.

[Copilot]

Pull request overview

Copilot reviewed 7 out of 7 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 4 changed files in this pull request and generated 4 comments.

[Copilot]

Pull request overview

Copilot reviewed 5 out of 6 changed files in this pull request and generated 1 comment.