Artifact ca1539a1e99461a07f99f376494e40a2d7951b1729c2cb14c010c71b03b57a39:

File conformance/rfcs/erratum/sync-protocol-01-s5.3-errors-count.md — part of check-in [8cc3d7c5e4] at 2026-10-13 02:23:50 on branch test/conformance-01-drafts — test(conformance): check the -01 drafts of artifact, repo-state and sync
Re-pin the conformance suite to draft-fossil-artifact-format-01, draft-fossil-repo-state-01 and draft-fossil-sync-protocol-01, copied byte for byte from zit check-in 100351e26c. The delta format has no -01 and stays on -00. The -00 errata are folded into -01 and are no longer pinned; the six -01 errata are pinned for the fossil 2.28 behavior they record. conformance/rfcs/README.md records where each copy came from.
The requirement tables are re-read against -01, and new cases cover the date tag's event time (repo-state §4.2 step 6), the Julian-day grammar, a delta manifest naming a delta, dropped reserved F cards, the trailing-punctuation retry, truncated payloads, errors N, a file card after a refused push, the end of a request after a failed login, and the answer to a gimme for private content.
No library behavior changes. Where -01 departs from fossil 2.28 and go-libfossil follows fossil (a cancelled, unparsable or 0 date tag), the cases are marked as deviations. The non-conformances the re-pin exposes are marked as gaps, tracked in #370, #371, #372, #373, #374 and #375. (user: [email protected], size: 2666) [annotate] [blame] [check-ins using] [more...]

Erratum — draft-fossil-sync-protocol-01 §5.3


Document	`draft-fossil-sync-protocol-01`
Section	5.3 — Comment Lines (the value of N in the timestamp comment)
Type	Technical (N is never defined; receivers are told not to act on it)
Status	Applied in `draft-fossil-sync-protocol-01` (2026-10-12)
Reported	2026-10-12
Submitter	Dan Mestas
Found via	A reader implementing §5.3 from the text alone

1. Summary

§5.3 gives the syntax of # timestamp DATETIME errors N and says N is informational, but never says what N counts. A sender has nothing to put there.

2. Location

§5.3, the paragraph beginning "DATETIME is exactly 19 bytes".

3. The reference behavior

Fossil 2.28 page_xfer() keeps a counter nErr (src/xfer.c:1278) that is incremented at each error card the server's own card handling emits: wrong project, the authorization refusals, multiple or failed login, the file, cfile and uvfile handler failures, and the script failures (:1330-1904). The closing comment prints it (:1970). The counter records errors raised, not cards present: a later reset discards the cards of earlier errors but not their count, so a push refused at :1517 followed by a file card refused at :1363 leaves one error card and errors 2. Three error cards are never counted: the unknown-operator bad command error (:1893), the private-content refusal emitted by server_private_xfer_not_authorized() (:1138-1140), and the SHA3 version refusal emitted by xfer_cannot_send_sha3_error() (:523-527), which is why Appendix A.2 shows that error followed by errors 0.

4. Impact

Cosmetic. N is informational and no receiver acts on it, but a cleanroom sender needs a definition.

5. Correction

    additional internal validation; DATETIME estimates skew, while N is
-   informational and not a wire error.
+   informational and not a wire error.  N counts the errors the
+   server raised while handling the request; it can exceed the error
+   cards present, because a reset discards earlier output.  Fossil's
+   server does not count the unknown-operator error, the private-
+   content refusal, or the SHA3 version refusal.  A receiver MUST NOT
+   act on N.

6. Rationale

Defines N as what Fossil's counter holds, records the three cards Fossil leaves out and the reset effect so a byte-comparing test is not surprised, and keeps N informational.

7. Verification

Source: Fossil 2.28 src/xfer.c:1278, every nErr++ between :1330 and :1904, :1893, :1138-1140, :523-527, :1970; Appendix A.2 of the draft.

Artifact [ca1539a1e9]