Fix fine-grained PAT attachment downloads for private repos (#477)

Fine-grained personal access tokens cannot download attachments from
private repositories directly due to a GitHub platform limitation.

This adds a workaround for image attachments (/assets/ URLs) using
GitHub's Markdown API to convert URLs to JWT-signed URLs that can be
downloaded without authentication.

Changes:
- Add get_jwt_signed_url_via_markdown_api() function
- Detect fine-grained token + private repo + /assets/ URL upfront
- Use JWT workaround for those cases, mark success with jwt_workaround flag
- Skip download with skipped_at when workaround fails
- Add startup warning when using --attachments with fine-grained tokens
- Document limitation in README (file attachments still fail)
- Add 6 unit tests for JWT workaround logic
This commit is contained in:
Rodos
2026-01-13 13:15:38 +11:00
parent c63fb37d30
commit fce4abb74a
4 changed files with 248 additions and 8 deletions

View File

@@ -281,6 +281,8 @@ The tool automatically extracts file extensions from HTTP headers to ensure file
**Repository filtering** for repo files/assets handles renamed and transferred repositories gracefully. URLs are included if they either match the current repository name directly, or redirect to it (e.g., ``willmcgugan/rich`` redirects to ``Textualize/rich`` after transfer).
**Fine-grained token limitation:** Due to a GitHub platform limitation, fine-grained personal access tokens (``github_pat_...``) cannot download attachments from private repositories directly. This affects both ``/assets/`` (images) and ``/files/`` (documents) URLs. The tool implements a workaround for image attachments using GitHub's Markdown API, which converts URLs to temporary JWT-signed URLs that can be downloaded. However, this workaround only works for images - document attachments (PDFs, text files, etc.) will fail with 404 errors when using fine-grained tokens on private repos. For full attachment support on private repositories, use a classic token (``-t``) instead of a fine-grained token (``-f``). See `#477 <https://github.com/josegonzalez/python-github-backup/issues/477>`_ for details.
Run in Docker container
-----------------------