Subj : Persist archive content listings in a per-area data/dirs/<code>.conten
To   : GitLab note in main/sbbs
From : Rob Swindell
Date : Sun Sep 20 2026 01:40:13

https://gitlab.synchro.net/main/sbbs/-/work_items/1247#note_10394

## Correction: the wasted ARC extraction was terminal-side, not web-side

The Problem section above says every view of a `.arc` file "re-attempts a doomed
extraction and returns an error page". That is true of the terminal viewer and
false of the web.

`webfileindex.ssjs` gates viewing on its own hardcoded list, which never
included the format:

```js
return ['zip', '7z', 'tgz', 'rar', 'lha', 'lzh', 'iso', 'cab'].indexOf(ext.toLowerCase()) >= 0;
```

So on the web those files were never offered a view link and a direct
`?view=` on one returned "Non-viewable file type" without reaching extraction
at all. The repeated doomed extraction happened through `[viewer:17]`, whose
`extension=*` catch-all hands every unmatched type to `archive.js`.

The waste was real, just on a different path than stated, and the fix has to
touch both: the web needs the format added to that list before the store's
fallback can ever be reached.

## Implementation status

Landed locally, not yet pushed:

- `exec/load/filecontents_lib.js` -- the store, path-to-area reverse mapping,
  `file_mutex` locking with a `js.on_exit` release, staleness, extraction with
  the external-tool fallback, and the three-tier read path.
- `exec/tests/filecontents_test.js` -- 30 checks.
- `exec/archive.js` -- `list` serves from the store; verbose still reads the
  archive, since CRC and compression format are not stored.
- `exec/webfileindex.ssjs` -- `view_archive()` serves from the store, entry
  names are HTML-encoded, and `arc` is added to the viewable list.

Verified on Vertrauen, on both consumers:

| path | result |
|---|---|
| terminal, live session | ARC files that previously said "Unsupported archive" now list, with correct color and alignment |
| web, `?view=` on a ZIP | renders, stored as `x: "libarchive"` |
| web, `?view=` on an ARC | renders, stored as `x: "lsar"`, written by the web server process |
| repeat requests | served from the store |
| concurrency | 4 processes, 100 locked writes to one area, store intact, no stranded locks |

The web result also settles the open question about allowing tier 2 there: the
web server successfully ran the external tool as the `sbbs` user and wrote the
store, so crawler traffic populates it as a side effect. Denying tier 2 on the
web was never really viable anyway, since it would have left `?view=` showing
nothing for any file the populator had not already reached.

Still open: `exec/inventory_archives.js` has not been rewritten and still
carries the `readd_always` import-date bug, so it should not be run.

-- *Authored by Claude (Claude Code), on behalf of @rswindell*
--- SBBSecho 3.37-Linux
 * Origin: Vertrauen - [vert/cvs/bbs].synchro.net (1:103/705)

.