mirror of
https://github.com/iv-org/invidious.git
synced 2026-08-16 21:10:53 -05:00
d7f990ebd7
Collaboration videos (multiple authors) have an extra metadataRows entry before the views/date one, containing the byline (e.g. "Channel A and Channel B"). The video branch of LockupViewModelParser always read metadataRows[0], which for these videos is the byline, not views/date, so both silently defaulted to 0 views and the current time. Fixed by searching every row's parts instead of assuming a fixed index, mirroring the same reasoning already used a few lines below for the playlist branch of this parser. Verified against real, live-captured YouTube API responses for both a single-author video and a real collaboration video (Veritasium x 2swap): confirmed the bug reproduces against the original code (views: 0) and is fixed by this change (views: 7900000). Full spec suite (174 examples) passes. Fixes: #5740 I want to be awarded the bounty associated to the issue this PR is fixing. -- AI disclosure (per AI_POLICY.md): this change was written with assistance from Claude Sonnet 5 (model id claude-sonnet-5), via Claude Code (Anthropic's CLI agent). The human (Alison Moura) read the AI usage policy, reviewed the diff, and reviewed the before/after test verification described above prior to this commit being made.