Owner decision 2026-10-16: the cans feature (bundles of pastes, #4) was
half-integrated and has zero prod usage; removing it eliminates four audit
findings by deletion.
Removed:
- internal/api/cans.go and all can routes (/api/pastes/can, /api/cans/*,
/can/{id}) plus handleCanPage/renderCan
- is_can from /api/mine and /api/public responses
- store layer: CanRow, can store funcs, paste_cans UNION in listings,
cans sweep; reserved slugs 'can'/'cans' released
- templates/can.html, can CSS, is_can badges in history.js/mine.js
- cans tests (cans_test.go, cans_flow_test.go, cans_expiry_test.go,
TestErrorCodeCanValidation)
- docs/api.md cans section, README/LICENSE/docker-compose mentions
Store migration: drops paste_cans only when empty (non-empty table is
kept and logged; prod verified clean 2026-10-16) and drops the dormant
pastes.can_id column. Covered by new TestMigrateDropsCansSchema.
Security hardening (audit findings): pw_<id> (unlock) and tok_<id>
(deletion-token) cookies now carry Secure: true in addition to
HttpOnly + SameSite=Lax; all deployments are HTTPS-only. cookie_set_test
now asserts Secure.
go vet ./... and go test ./... pass.
- history/mine table column 'Language' -> 'Type' (data-sort key 'type')
- /api/public and /api/mine rows gain a 'type' field: file extension for
attachment pastes, stored language otherwise (default 'text')
- paste page stats label 'Language' -> 'Type'; summary line uses the same
label; get-paste JSON gains 'type'
- store list queries LEFT JOIN attachments to expose the filename
- table.js sorting accepts the 'type' key
SoftDelete now reports whether it performed the delete (conditional
UPDATE ... WHERE deleted_at IS NULL checked via RowsAffected).
RegisterRead returns an admitted flag: legacy burn pastes admit exactly
one reader (the atomic soft-delete winner), and burn-after-N pastes
increment reads_used via a conditional UPDATE guarded on
reads_used < reads_limit, so concurrent readers cannot both consume the
final read. API, HTML, and raw read paths return 404 when the reader
loses the burn claim; content is never served twice.
OpenStore pins the SQLite pool to one connection: concurrent writes on
separate pooled connections surfaced SQLITE_BUSY as spurious 500s
instead of serializing.
Adds concurrency regression tests: 24 parallel readers of a burn paste
(exactly one receives content, none of the others leak it) and 30
parallel readers vs a 3-read budget (exactly 3 admitted, then 404).