Add Midra aspect mode operator test checklist
Beffroi/beffroi-tools/pipeline/head There was a failure building this commit Details

This commit is contained in:
Dandaedre 2026-11-15 17:42:40 +01:00
parent 796b16d36f
commit ab073e876b
3 changed files with 187 additions and 1 deletions

View File

@ -4,6 +4,9 @@
## Unreleased
- Add an operator checklist and results template for Midra fill/aspect tests,
covering scaling, inheritance, crop order, input correction and rounding.
- Prioritize Midra input layer-fill and layer aspect support. Review the official
V3.2 user manual, AWJ v3.2 and REST v3.0.13 guides; record the remaining
inheritance, scaling, cropping and physical-output checks in the test list.

View File

@ -0,0 +1,182 @@
# Midra fill and aspect mode test checklist
Purpose: determine the scaling formulas, setting precedence, crop order and
clipping behavior for input **Layer fill** and layer **Aspect** modes.
Start with sections A and B and share the results before proceeding. They may
let us shorten sections C and D. Untested cases should stay unchecked.
## Reference setup
- [ ] Record the Midra model and firmware.
- [ ] Save the current device configuration.
- [ ] Use a **1920 × 1080 source** with a visible outer border, center crosshair,
circle and numbered grid. This makes stretching and missing edges obvious.
- [ ] Use a **1920 × 1080 output**.
- [ ] Set one live layer to **800 × 600**, centered at **(960, 540)**.
Its top-left should be **(560, 240)**.
- [ ] Disable masks, borders, shadows and other effects.
- [ ] Set input crop and layer crop to zero.
- [ ] Set input **Content ratio = Native**, **Transform to = Signal Aspect Ratio**,
and **Layer fill = Follow layer settings**.
- [ ] Change settings directly in WebRCS. **Do not send from Canvas Pilot during
these tests:** its current geometry sends force Centered.
Judge rendering using physical output or its multiviewer image. WebRCS alone
may show only the container. If measuring from a scaled multiviewer screenshot,
record its scale or include the full output frame so coordinates can be derived.
Keep output clipping distinct from clipping at the layer boundary.
Copy this record for each test. Keep screenshots named after the test ID.
If a measurement is unavailable, record what is visible rather than guessing.
```text
Test ID:
Model / firmware:
Source signal dimensions:
Input Content ratio / Transform to / Layer fill:
Input crop (left, right, top, bottom; units):
Layer Aspect:
Layer center and size:
Layer crop (left, right, top, bottom; units):
Preview or Program:
Visible image bounds (left, top, right, bottom; output pixels):
Missing source edges / surviving grid lines:
Circle round or stretched:
Reported content dimensions, if displayed:
Screenshot filename:
Other observations / settings automatically changed:
```
## A. Basic scaling modes
Keep the reference setup unchanged except for **layer Aspect**.
| Done | Test | Layer Aspect |
| --- | --- | --- |
| [ ] | A1 | Centered |
| [ ] | A2 | Fullscreen |
| [ ] | A3 | Cropped |
| [ ] | A4 | 1:1 |
For each, record visible bounds, whether the entire source border survives,
whether the circle stays round, and whether content extends outside the layer.
- [ ] **A5:** Repeat 1:1 with an actual **640 × 360 source**, retaining the
800 × 600 layer. Record image placement. Change the actual signal resolution,
not just the image drawn within a 1920 × 1080 signal.
- [ ] **A6a:** Restore the 1920 × 1080 source. Use a **600 × 800 layer**, still
centered at (960, 540), with Centered aspect.
- [ ] **A6b:** Repeat A6a with Cropped aspect.
## B. Input/layer precedence and inheritance
Restore the 800 × 600 layer and zero crop. Keep Content ratio = Native and
Transform to = Signal Aspect Ratio.
| Done | Test | Input Layer fill | Layer Aspect |
| --- | --- | --- | --- |
| [ ] | B1 | Centered | Input setting |
| [ ] | B2 | Fullscreen | Input setting |
| [ ] | B3 | Cropped | Input setting |
| [ ] | B4 | 1:1 | Input setting |
| [ ] | B5 | Fullscreen | Centered |
| [ ] | B6 | Centered | Fullscreen |
| [ ] | B7 | Follow layer settings | Input setting |
- [ ] Compare B1–B4 with A1–A4.
- [ ] Record which setting wins in B5 and B6.
- [ ] Record B7's result, including any automatic setting change or refusal.
- [ ] **B8:** Layer Aspect = Global settings; input fill = Follow layer settings.
Change the global mode between Centered and Fullscreen. Record both results
and where the global setting is located.
- [ ] **B9:** Repeat B8 with input fill = Cropped.
- [ ] **B10:** Repeat B7 while changing the global mode between Centered and
Fullscreen. This tests whether global mode resolves the inheritance loop.
**Share A and B before continuing.**
## C. Crop order
Restore the 1920 × 1080 source, Native / Signal Aspect Ratio, input fill = Follow
layer settings, and the 800 × 600 layer centered at (960, 540). Reset input and
layer crop before each case, then apply only the values listed.
| Group | Input crop | Layer crop |
| --- | --- | --- |
| C1 | Zero | Left 25%; other edges zero |
| C2 | Left 480 source pixels; other edges zero | Zero |
| C3 | Left 480 source pixels; other edges zero | Top 25%; other edges zero |
Run each group in all four layer modes:
- [ ] C1-Centered
- [ ] C1-Fullscreen
- [ ] C1-Cropped
- [ ] C1-1:1
- [ ] C2-Centered
- [ ] C2-Fullscreen
- [ ] C2-Cropped
- [ ] C2-1:1
- [ ] C3-Centered
- [ ] C3-Fullscreen
- [ ] C3-Cropped
- [ ] C3-1:1
Record surviving grid lines, the source center's position, visible bounds and
reported content dimensions. Note whether retained content is recentered and
whether WebRCS changes displayed crop values or units. Capture the values after
applying them, including any rounding.
## D. Aspect-correction order
Reset all crop. Restore the reference source/layer sizes and input fill = Follow
layer settings. Reset corrections between cases before applying this table.
| Test | Content ratio | Transform to |
| --- | --- | --- |
| D1 | 4:3 | Signal Aspect Ratio |
| D2 | Native | 4:3 |
| D3 | 4:3 | 16:9 |
- [ ] D1-Centered
- [ ] D1-1:1
- [ ] D2-Centered
- [ ] D2-1:1
- [ ] D3-Centered
- [ ] D3-1:1
- [ ] D4-Centered: repeat D2 with 480 pixels of left input crop.
- [ ] D4-1:1: repeat D2 with 480 pixels of left input crop.
Record reported content dimensions as well as the rendered result. These tests
help distinguish pixel-aspect interpretation from effective image dimensions.
## E. Consistency and restoration
- [ ] **E1:** Repeat one distinctive cropped case in Preview and Program. Record
the case ID and confirm the settings are otherwise identical.
- [ ] **E2:** Put the same input into two layers: one with explicit Centered,
the other with Input setting. Change input fill from Centered to Fullscreen
and record which layer changes. Keep Content ratio and Transform to at defaults.
- [ ] **E3:** With default input settings and zero crop, try an **801 × 601**
layer at (960, 540) using Centered, then Cropped, to expose rounding.
- [ ] Restore the saved configuration.
## Results and conclusions
Fill this in after comparing observations; formulas remain hypotheses until
supported by the recorded output.
- Effective mode precedence:
- Global/input inheritance fallback:
- Centered scaling and placement:
- Fullscreen scaling:
- Cropped scaling and clipping:
- 1:1 placement and clipping:
- Input correction / input crop / layer crop processing order:
- Meaning of reported content dimensions:
- Rounding behavior:
- Remaining ambiguous cases:
Related: [project test list](plan.md#remaining-tests) and
[current Midra send contract](../interfaces/midra-device-send.md).

View File

@ -125,7 +125,8 @@ integration planned for this scope.
### Remaining tests
- [ ] **Midra fill/aspect acceptance.** Cross input fill modes with explicit,
- [ ] **Midra fill/aspect acceptance.** Follow the
[operator test checklist](midra-aspect-test-checklist.md). Cross input fill modes with explicit,
input-inherited and global layer modes; include conflicting settings and both
inheritance defaults. Use mismatched source/container ratios, smaller/larger
1:1 sources, asymmetric input/layer crop, content-ratio correction and Transform