tracker

5.3 KB

1. # Report 026 (w075) — tracker#31 duplicate titles + deploy.sh waits for the boot
3. **Commit**: `dc40d85` on main in loreana:/media/STORAGE/projects/tracker.worldapi.org (parent 2667da0). Not pushed, not deployed.
4. I never read `.env`. Every process was killed by its PID, and nothing of mine is still running. The real-data copies are deleted.
6. ## Cause
7. - **The duplicate titles come from the OLD tracker, not from this app.**
8. - Live copy of 2026-10-03: 68 type+tmdbId pairs are held by 157 records.
9. - Every one of the 157 is a migrated row (Mongo `oldId`). They were made in 2022–2023 as runs of consecutive ids, and several share one slug (Archive 81 ×6 → 5 × `Archive-82`).
10. - No record made by this app is a duplicate. Its two creating paths already look the title up first:
11. - the search import (`search.hl importTitle`);
12. - the filmography fill (`people.hl fillStep` → `titleIdByTmdb`).
13. - Their records carry `oldId tmdb-<tv|movie>-<id>`, which the `!oldId` UNIQUE index refuses a second time.
14. - **Seasons**: 449 migrated seasons name a title that does not list them (the old tracker showed seasons by query).
15. - The daily sync goes by the list, so where TMDB has that season number it made a twin (16 such).
16. - Reacher: the creator's 8 S3 watches + the season check sat on the unlisted migrated S3. The page showed the sync's S3 as unwatched.
18. ## What was done
19. - **Creating paths**:
20. - The import looks the id up again right before its put.
21. - A merged tombstone answers its keeper everywhere (`shows.hl liveShowOf`, `search.hl searchTitleMerged` points the TMDB id at the keeper).
22. - No new index: an mpackdb unique index refuses a second null (tombstones, titles without a TMDB id), and adding an index to an existing table is untested (hybriel#21).
23. - Gate check per path: `tests/paths-m026.hl`.
24. - **Repair** (`merge.hl`, project.hl `mergeTick`): runs in the app on its own 0.1 s clock, one item per tick, no TMDB. The TMDB jobs wait for it. It resumes after a restart; later starts log `merge: nothing to do`.
25. - **Keeper**: the record with follows/watches, else the one with its old short id, else the oldest.
26. - **Moved to the keeper**: follows, watches, seasons (moved, or merged episode by episode), cast + crew, persons' credits, timelines and empty fields.
27. - **Tombstones**: the others get `mergedInto`, `mergedTmdbId`, `mergedSeasons`. Nothing is deleted. A repeated watch/follow is parked (`mergedInto`, target null).
28. - **Slugs**: the keeper takes its plain slug when a duplicate held it (Archive 81: `Archive-82` → `Archive-81`).
29. - **Redirects**: tombstone slugs and keeper former slugs answer 301 (function routes built at boot; `/shows/` via showsRoute). Tombstone short ids answer 301 at once.
30. - **Stray seasons**: a stray with a listed twin is merged into it. A stray without a twin is linked only when someone watched it.
31. - **deploy.sh**: step 5 asks every 2 s for up to 90 s (`DEPLOY_URL_WAIT`) and prints only changes. Tested against a fake server: 502 for 12 s → `after 12 s: 200` (exit 0); with a 6 s limit → REFUSED.
33. ## Real data (copy of live `storage/`, TMDB jobs off; before = 2667da0 on the same copy)
34. | | before | after |
35. |---|---|---|
36. | count-tmdb-ids: type+tmdbId shared | 68 ids | **0** (89 tombstones) |
37. | Archive 81: live records / search hits | 6 / 6 | **1 / 1**, `/series/Archive-81`, all 8 checked |
38. | Reacher listed S3 watched / check | 0 of 8 / regular | **8 of 8 / solid** |
39. | /my/series Reacher | S03E08 | S03E08 |
40. | /my/unwatched rows | 1,273 (incl. Reacher S03) | 1,265 (Reacher S04 only) |
41. | az5b2 follows rows/live, watches rows/live | 130/130, 12,139/12,136 | 130/130, 12,139/12,136 |
43. - **Merge log**: `merge done: groups=68 tombstones=89 watches=9 seasonsMerged=29 seasonsLinked=2 episodesMoved=22 episodesMerged=187 cast=16 people=44 slugs=1 … slowestStepMs=187 seconds=9.869`.
44. - **Next start**: `moved slugs: 28`, `merge: nothing to do`.
45. - **curl**: `/series/Archive-82` and `/727vk` → 301 `/series/Archive-81`; `/movies/Ecstasy-10` → 301 `/movies/Ecstasy-9`.
46. - **Boot to listening**: 24 s → 26–27 s.
47. - **Screenshots** (looked at): `/tmp/m026/real/{before,after}-*-{390,1280}.png` on Loreana.
49. ## Gates (ports 8720–8727)
50. - `tests/browser.mjs` → `342 passed, 0 failed` (+15 #31 checks).
51. - `tests/kinds.mjs` → `32 passed, 0 failed`.
52. - `tests/franchises.mjs` → `52 passed, 0 failed`.
53. - Shots looked at: `/tmp/m026/shots/dupe-{keeper,stray}-1280.png`.
55. ## Open points
56. 1. **Unlinked stray seasons (decision for the architect)**: 431 stray seasons stay unlinked (391 have episodes, none watched, no twin).
57. - My first run linked all of them. That added 1,091 never-watched old talk-show episodes to the creator's /my/unwatched, so I stopped it.
58. - Should they be linked anyway?
59. 2. **Old slugs during the first run**: a slug merged while the process runs shows the keeper's page, but answers 301 only from the next start. The deploy note says to restart once after `merge done`.
60. 3. **Duplicate episode numbers**: 224 migrated episodes share season+number inside one season. These were not touched.
61. 4. **Scan on every start**: `mergePrepare` scans shows/seasons/watches on the first tick of every start (~1.2 s, blocking requests once).
63. ## Hybriel
64. No new bugs found.
65. - **Note for the workers' guide**: the manifest's `routes` must stay an array LITERAL, because hl:web reads its syntax to find the page modules. Extra routes go into the constructor's `routes =` (here `allRoutes`).