|
2 | 2 |
|
3 | 3 | Release history for the DaVinci Resolve MCP Server. The latest release is summarized in the root README; older entries live here to keep the README focused. |
4 | 4 |
|
| 5 | +## What's New in v2.72.1 |
| 6 | + |
| 7 | +Documentation only. The API coverage page stated its method counts in four |
| 8 | +places and they disagreed; the cause turned out to be structural rather than |
| 9 | +clerical. |
| 10 | + |
| 11 | +### The Resolve 21 surface was never counted |
| 12 | + |
| 13 | +None of the Resolve 21 methods appeared in the Complete API Reference tables — |
| 14 | +the tables the `API Methods Covered` denominator counts — although the server |
| 15 | +has wrapped, released and live-tested them since v2.28.1. So `337/337 (100%)` |
| 16 | +described a surface that excluded nine methods across four classes, and the |
| 17 | +`336 → 337` bump for `ResetIntellisearchAnalysis` added one to a count whose |
| 18 | +table did not list it. |
| 19 | + |
| 20 | +Thirteen rows added, one per method per object class, since a wrapper on |
| 21 | +`Folder` and one on `MediaPoolItem` can fail independently. Signatures taken |
| 22 | +from the bundled 21.0.2 scripting reference. `TranscribeAudio` was already |
| 23 | +listed — its Resolve 21 change is the optional `useSpeakerDetection` argument, |
| 24 | +accepted but producing identical transcripts either way — so those rows are |
| 25 | +annotated rather than duplicated. |
| 26 | + |
| 27 | +Every summary figure is now derived from the tables: **349 covered, 338 live |
| 28 | +tested, 11 untested**. |
| 29 | + |
| 30 | +### The counting convention is now written down |
| 31 | + |
| 32 | +Two of the three disagreements came from it being implicit: |
| 33 | + |
| 34 | +- A method that could not be executed is **not** counted as a pass. The old |
| 35 | + "Resolve 21 delta 8/9" counted Extras-blocked methods as passes, contradicting |
| 36 | + the prose directly above it and overstating coverage exactly where the risk is |
| 37 | + highest. |
| 38 | +- The phase table counts **methods, not assertions**, which is why its Total |
| 39 | + equals Methods Live Tested. That ambiguity is what made two figures look |
| 40 | + independently wrong. |
| 41 | + |
| 42 | +`tests/test_api_coverage_arithmetic.py` derives all four figures from the |
| 43 | +reference tables and fails if any disagrees, so the tables stay the single |
| 44 | +source. It was checked against both drift shapes: a hand-edited summary figure, |
| 45 | +and a row removed from a table. |
| 46 | + |
5 | 47 | ## What's New in v2.72.0 |
6 | 48 |
|
7 | 49 | Resolve 21's AI methods report a missing Extras pack as an error *string*, not |
|
0 commit comments