fix(vdisplay): recommit by the monitor's CCD key, not a bare target id #562

Merged

enricobuehler merged 1 commits from fix/vdisplay-gdi-name-key into main 2026-09-02 21:56:42 +00:00

Conversation 0 Commits 1 Files Changed 1

+17 -3

enricobuehler commented 2026-09-02 19:20:44 +00:00

Owner

Copy Link

Copy Source

windows-host.yml fails at "Build (release, nvenc + amf-qsv + qsv)" on main since #555 met the win-display actor work:

error[E0308]: mismatched types
   --> crates\pf-vdisplay\src\vdisplay\windows\manager.rs:527:67
527 |     let Some(gdi) = pf_win_display::win_display::resolve_gdi_name(target_id) else {
    |                                                                   ^^^^^^^^^ expected `CcdTargetKey`, found `u32`

Two merges crossed: resolve_gdi_name now takes a CcdTargetKey (target ids are only unique per adapter), while the UAC mid-stream fix's force_recommit(target_id: u32) still passed the bare id. Windows-only code, so the Linux lanes never compiled it.

Fix: look the monitor up in the manager's slots under the state lock (the lock was already held there — sole topology mutator), use its cached GDI name, or resolve it by its own ccd_key(); a target the manager does not own is a no-op. The caller in punktfunk-host/src/capture.rs keeps its u32 signature.

Verified: cargo check -p pf-vdisplay --release on the .133 runner at this commit, Finished. The full windows-host.yml run on main after merge is the real proof.

`windows-host.yml` fails at "Build (release, nvenc + amf-qsv + qsv)" on main since #555 met the win-display actor work: ``` error[E0308]: mismatched types --> crates\pf-vdisplay\src\vdisplay\windows\manager.rs:527:67 527 | let Some(gdi) = pf_win_display::win_display::resolve_gdi_name(target_id) else { | ^^^^^^^^^ expected `CcdTargetKey`, found `u32` ``` Two merges crossed: `resolve_gdi_name` now takes a `CcdTargetKey` (target ids are only unique per adapter), while the UAC mid-stream fix's `force_recommit(target_id: u32)` still passed the bare id. Windows-only code, so the Linux lanes never compiled it. Fix: look the monitor up in the manager's slots under the state lock (the lock was already held there — sole topology mutator), use its cached GDI name, or resolve it by its own `ccd_key()`; a target the manager does not own is a no-op. The caller in `punktfunk-host/src/capture.rs` keeps its `u32` signature. Verified: `cargo check -p pf-vdisplay --release` on the `.133` runner at this commit, `Finished`. The full `windows-host.yml` run on main after merge is the real proof.

enricobuehler added 1 commit 2026-09-02 19:20:45 +00:00

fix(vdisplay): recommit by the monitor's CCD key, not a bare target id ... f6f69547f1

Two merges crossed: pf-win-display's resolve_gdi_name now takes a
CcdTargetKey (target ids are only unique per adapter), while the UAC
mid-stream fix's force_recommit still passed a u32. Windows-only code, so
only windows-host.yml's build step saw it. Look the monitor up in the
manager's slots under the state lock, use its cached GDI name, or resolve it
by its own key; a target the manager does not own is a no-op.

enricobuehler referenced this pull request 2026-09-02 19:22:04 +00:00

fix(host): keep handshake's test module last #563

enricobuehler merged commit 15cb7931ac into main 2026-09-02 21:56:42 +00:00

enricobuehler deleted branch fix/vdisplay-gdi-name-key 2026-09-02 21:56:43 +00:00

Sign in to join this conversation.

Reviewers

No Reviewers

Labels

No items

No labels

Milestone

No items

No Milestone

Projects

Clear projects

No projects

Assignees

Clear assignees

FernandoMX FlipperDan HitFrostbite Seenko bluelightspecial enricobuehler (Enrico Bühler) kszlim luxus nannku neoney sonar

No Assignees

1 Participants

Notifications

Subscribe

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: unom/punktfunk#562