On Tue Apr 21 18:44:52 2026 +0000, Jan Robert Gerigk wrote:
Rebased so each test sits next to its source commit. I also did a second review pass on the three new tests before pushing: 2/10 DIGCF_DEVICEINTERFACE filter: uses device ID 0005 to stay clear of other tests using 0000-0003, uppercase-strcmp for the instance ID comparison (matches existing convention in the file). 6/10 DEVPKEY_Device_Children: fetches the bus device via bus_class + DIGCF_DEVICEINTERFACE | DIGCF_PRESENT, matching the enumeration pattern used elsewhere in test_pnp_devices. 8/10 DEVPKEY_Device_Siblings: transient second child; counters and the DevQuery binary semaphores (device_added_sem / device_removed_sem) are drained on exit so the later WAIT_TIMEOUT assertions on those semaphores remain valid. Local setupapi_test devinst run: 4376 tests, 0 failures. ntoskrnl test compiles clean Sorry for the back-and-forth across multiple review rounds, hopefully this one is now fine. One question: I saw a merge attempt last Friday that did not go through because the branch was 49 (?) commits behind master. The series is now rebased onto wine 11.7. Does the merge pool automatically retry on the next merge window, or is there something I should do? Pipeline showed 13 failures from my last push. Analysed locally:
6/10 (now 6/9): my "cleaner" SetupDi pattern from the previous review pass was wrong for Wine. SetupDiEnumDeviceInfo on a DIGCF_DEVICEINTERFACE set returned no items. Fixed by switching to SetupDiEnumDeviceInterfaces + SetupDiGetDeviceInterfaceDetail-with-NULL-detail to extract the bus device's SP_DEVINFO_DATA. Setupapi devinst test still 0 failures locally. 8/10 (the Siblings test): the transient second-child approach broke kernel-side counters in driver_pnp.c (hardcoded remove_device_count == 0) and the user-mode notification handlers (hardcoded "Wine#Test #1" strings). I dropped the test from this MR. The Siblings property storage commit (now 7/9) stays in. If that is OK with you, I would submit a dedicated Siblings test as a follow-up MR with its own bus-test fixture so the existing test_pnp_devices invariants stay intact? Series rebased onto current master (was 34 commits behind), 9 commits total now. Running some tests again locally, should by ready any time soon. -- https://gitlab.winehq.org/wine/wine/-/merge_requests/10604#note_137253