Discovered while reviewing the #1404 rename logs (2.29.0 dry-run / 2.29.1 rollout). Seven Talk rows reference files that no longer exist on disk, so on every container start they show up as Re-standardized churn in restandardize_artifact_filenames (the rename no-ops on a missing source) and as repair_diverged_artifact_filenames warnings ("no matching orphan was found on disk; left untouched for manual review").
The now-live #1390 repair (promoted in 2.29.1) already recovered the 3 files that had identifiable orphans (pdfs of ids 528, 564, 513). The remainder are genuinely missing — no orphan candidate exists:
| Talk id |
Missing field |
Missing file |
| 638 |
pdf_file |
talks/Froehlich_MakingWithASocialPurpose_2017.pdf |
| 639 |
pdf_file |
talks/Froehlich_MakingWithASocialPurpose_2017.pdf |
| 665 |
pdf_file |
talks/Jain_SoundSensingAndFeedbackTechniquesForDDeafAndHardOfHearingUsers_2021.pdf |
| 528 |
raw_file |
talks/Sensing_and_Predicting_the_Pulse_of_the_City_through_Shared_Bicycling.pptx |
| 564 |
raw_file |
talks/Infrastructure_Mediated_Sensing_for_Assistive_Environments.pptx |
| 571 |
raw_file |
talks/UbiGreen_Investigating_a_Mobile_Tool_for_Tracking_and_Supporting_Green_Transportation_Habits.pptx |
| 513 |
raw_file |
talks/Social_Fabrics_Designing_Wearable_Electronics_Textiles_for_Interaction_Introspection_and_Inquiry_Ln1NPpe.pptx |
(638 and 639 are two 2017 "Making with a Social Purpose" talks — DUB Retreat and UMD CS Staff Talk — whose DB rows point at the same filename, which suggests one upload clobbered the other before uniquification existed.)
To resolve, per row
- Re-upload the file from a personal archive if it exists (best outcome), or
- Clear the dead FileField in
/admin so the row stops referencing a nonexistent file.
Either action ends the churn for that row: once the field is empty or points at a real file, restandardize_artifact_filenames and repair_diverged_artifact_filenames both go quiet.
Where the evidence lives
🤖 Generated with Claude Code
Discovered while reviewing the #1404 rename logs (2.29.0 dry-run / 2.29.1 rollout). Seven Talk rows reference files that no longer exist on disk, so on every container start they show up as
Re-standardizedchurn inrestandardize_artifact_filenames(the rename no-ops on a missing source) and asrepair_diverged_artifact_filenameswarnings ("no matching orphan was found on disk; left untouched for manual review").The now-live #1390 repair (promoted in 2.29.1) already recovered the 3 files that had identifiable orphans (pdfs of ids 528, 564, 513). The remainder are genuinely missing — no orphan candidate exists:
talks/Froehlich_MakingWithASocialPurpose_2017.pdftalks/Froehlich_MakingWithASocialPurpose_2017.pdftalks/Jain_SoundSensingAndFeedbackTechniquesForDDeafAndHardOfHearingUsers_2021.pdftalks/Sensing_and_Predicting_the_Pulse_of_the_City_through_Shared_Bicycling.pptxtalks/Infrastructure_Mediated_Sensing_for_Assistive_Environments.pptxtalks/UbiGreen_Investigating_a_Mobile_Tool_for_Tracking_and_Supporting_Green_Transportation_Habits.pptxtalks/Social_Fabrics_Designing_Wearable_Electronics_Textiles_for_Interaction_Introspection_and_Inquiry_Ln1NPpe.pptx(638 and 639 are two 2017 "Making with a Social Purpose" talks — DUB Retreat and UMD CS Staff Talk — whose DB rows point at the same filename, which suggests one upload clobbered the other before uniquification existed.)
To resolve, per row
/adminso the row stops referencing a nonexistent file.Either action ends the churn for that row: once the field is empty or points at a real file,
restandardize_artifact_filenamesandrepair_diverged_artifact_filenamesboth go quiet.Where the evidence lives
debug.log(viarecycle/makelab1:/cse/web/research/makelab/www/debug.log), any container start:Re-standardized Talk id=...lines for these ids +repair_divergedWARNINGs.🤖 Generated with Claude Code