Handle corrupt GF3 archives in batch runs

This commit is contained in:
GF3 Pipeline Bot
2026-05-30 01:46:04 +08:00
parent 5c24c0e64c
commit ff41cf3ced
6 changed files with 125 additions and 9 deletions
+16
View File
@@ -40,6 +40,22 @@ The wrapper supports:
Directory mode recursively scans for `*.tar.gz`, `*.tgz`, and `*.meta.xml`. If both archive and metadata exist for the same scene name, the archive is preferred. Internal `.gf3_extract` folders are skipped.
## Batch Failure Handling
Directory runs are fault-tolerant at the scene level. If a single scene cannot be extracted or processed, the wrapper records that scene as failed, prints the input path and error reason, and continues with the next scene.
Archive read errors include the archive path, archive size when available, and the archive member being read when the failure happened. For example, a truncated `.tar.gz` reports that the archive appears incomplete and should be re-copied or re-downloaded.
At the end of the run, the wrapper prints a summary:
```text
Batch summary: 3 succeeded, 1 failed, 4 total
Failed scenes:
- GF3_... (<archive path>): <failure reason>
```
If any scene failed, the process exits with a non-zero status after the summary. This lets scheduled jobs detect that the batch needs attention while still preserving successful outputs from other scenes.
## Outputs
Each scene is processed into its own output directory: