Fossil Anomaly and Example Seeds

Use these as review seeds and example-capture traps. They are starting points for source or observed-byte checking, not facts to promote blindly.

Artifact-focused seeds

Seed	Why it matters	Check against
A wiki artifact created by append with submitting user nobody may carry no U card.	This is the one explicit known case named in the original prompt; it tests whether the reader rule differs from a writer quirk.	src/manifest.c parser behavior plus an observed artifact if one exists
Technote N-after-P ordering relaxation.	The relaxation is legitimate only if the technote type admits P; otherwise a reader may mistake it for an editorial error.	card-order logic in src/manifest.c and any observed technote instances
fossil test-parse-manifest may reject an artifact that manifest_parse admits.	The command's pre-check is narrower than the library parser, so anomaly judgments must resolve to the library rule.	manifest_is_well_formed versus manifest_parse
Delta-manifest F card with no hash means delete, not empty content.	Readers will misbuild delta-manifest file sets if this is left implicit.	delta-manifest merge path in src/manifest.c plus observed delta manifests
Existing SHA-1 artifacts remain readable.	Historical repository readability is a document requirement, not a legacy footnote to drop.	naming/hash-policy handling plus observed SHA-1 artifacts

Sync-focused seeds

Seed	Why it matters	Check against
The client POSTs to the repository URL path unchanged; it does not append /xfer.	This is a high-risk interoperability bug if specified wrongly.	POST construction in src/http.c and sync dispatch in src/xfer.c
file / cfile payloads are followed immediately by the next operator, while config payload is followed by a newline the receiver skips.	Payload framing is not uniform across sync cards.	send/receive paths in src/xfer.c and src/configure.c
A delta-form file or cfile may arrive before its source during clone and must be retained until the source arrives.	Rejecting or eagerly failing this case breaks clone correctness.	content_xfer, xfer_accept_file, and dangling-delta handling in src/xfer.c
Current clone path is cfile plus clone_seqno; older send_file clone path is legacy and out of scope.	Prevents mixing current and obsolete wire versions in one document.	current clone path in src/xfer.c, plus version-change handling via version pragmas
Pre-2.27 login handling is out of scope, but the actual version/date boundary still has to be stated from source.	Avoids vague historical cut lines.	login/version handling in src/xfer.c

Delta- and example-focused seeds

Seed	Why it matters	Check against
The header newline in a delta is structural and must remain visible in rendered examples.	Rendered examples can accidentally collapse the newline and misstate the format.	observed create/apply examples and rendered text extraction
Literal payloads may contain 0x0a and must be located by count, not by scanning.	A line-splitting reader will mis-parse valid deltas.	delta parser/apply paths plus observed examples containing embedded newlines
Source-absent behavior must be taken from implementation, not from generic delta expectations.	The suite promises implemented behavior, not generic delta folklore.	src/delta.c, src/deltacmd.c, src/deltafunc.c

Example-rendering traps from the original pack

These are not protocol anomalies; they are recurring failures in rendered RFC examples.

• Invisible bytes: leading/trailing spaces, tabs, or repeated spaces disappear in prose unless the example exposes them with byte lengths or explicit delimiters.
• Collapsed structure: a required separator byte, especially 0x0a, must be shown visibly rather than running into the next token.
• Unstated field encoding: if an example shows a date, time, size, hash, version, or sequence number, the running text must state its encoding so the reader can reproduce it.
• Observed, never predicted: examples must come from implementation output and must round-trip from rendered text back through the implementation.

Promotion rule

A seed graduates into a public anomaly appendix entry only after source or observed-byte checking shows that the behavior is real and the public document needs to preserve or explain it.