test(extra): cover proxy pipe cross-process access

[Marc-André Moreau (mamoreau-devolutions)]

Summary

• verify a DvcNamedPipeProxy Windows pipe is visible to a separately spawned process
• require the child process to forward pipe data into the parent DVC callback

Tests

• cargo test -p ironrdp-testsuite-extra --test integration_tests_extra dvc_pipe_proxy
• cargo clippy -p ironrdp-testsuite-extra --test integration_tests_extra --no-deps -- -D warnings

[Copilot]

Pull request overview

Adds Windows integration coverage proving DvcNamedPipeProxy supports cross-process pipe access and forwards data into the DVC callback.

Changes:

• Spawns a separate test process as a named-pipe client.
• Refactors shared proxy setup and forwarding assertions.
• Adds the required ironrdp-svc test dependency.

Reviewed changes

Copilot reviewed 2 out of 3 changed files in this pull request and generated no comments.

File	Description
crates/ironrdp-testsuite-extra/tests/dvc_pipe_proxy.rs	Adds cross-process pipe forwarding coverage.
crates/ironrdp-testsuite-extra/Cargo.toml	Adds the SVC test dependency.
Cargo.lock	Records the dependency update.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Benoît Cortier (CBenoit)]

Thank you!

Some review items from Claude that I think are worth addressing:

The hardcoded child-test path is stale-prone with a misleading failure mode

dvc_pipe_proxy.rs:48 — "dvc_pipe_proxy::child_process_connects_and_forwards_windows_pipe_data"
must track both the module name in main.rs (mod dvc_pipe_proxy;) and the fn name. If
either is renamed, --exact matches zero tests; libtest then exits 0, so
assert!(status.success()) (line 54) passes, and the test instead fails 1s later at
assert_forwarded_data with "must forward pipe data to its callback" — pointing at a
forwarding bug that doesn't exist. This is the first use of the current_exe re-exec
pattern in the repo, so there's no established guard. Consider asserting the child
actually ran a test (e.g. parse --list or check the summary), or at minimum a comment
binding the string to the fn so a rename is caught in review.

child_process_… is a real test that silently no-ops in normal runs

dvc_pipe_proxy.rs:62-65 — as a #[tokio::test] it's discovered in every normal run,
where the env var is absent so it returns Ok early and shows up green. Its name reads
like a standalone assertion ("connects_and_forwards…") but in the normal suite it
asserts nothing. A name that signals its helper role (e.g. child_process_helper_…)
would make the test output honest without changing the --exact filter meaningfully.