epiphany/crates/epiphany-testkit/tests/binary_format_history.rs

315 lines
13 KiB
Rust

//! A scoped guard on `spec/binary_format.tex`'s Revision History chapter
//! (Packet B of `spec/CONTRACT_GENESIS_G3A_UNDO_REPAIR.md`, pins B5/B6),
//! filed against P13-S17: the chapter once ran genesis tranche G2a straight
//! to G-minor to G3a, omitting G2b entirely — including the accept-set
//! raise `OperationEnvelopeBlock` 2 -> 3 that G2b performed, which reached
//! the normative tables but never the history.
//!
//! Two constraints from pin B6 pull against each other and both must hold:
//!
//! * Bare name-presence is not enough. `"G2b"` already occurs inside the
//! chapter in prose — the G3a row observes that the accept-set "stays at
//! 3 where genesis tranche G2b left it" — so a guard that only checked
//! for the substring `"G2b"` would stay green even with the G2b row
//! deleted. The guard below requires a *principal marker*: the rung name
//! immediately preceded by the row's `---` separator (e.g.
//! `--- Genesis tranche G2b`), which a prose mention where the separator
//! *follows* the name (as in "genesis tranche G1 --- landed at",
//! `binary_format.tex:3603`) cannot satisfy.
//! * No document version number appears anywhere in this file — not in the
//! assertions and not in this comment, which is why none is quoted here
//! even as an example. Encoding one would pin a number a future chronology
//! correction would have to move, reintroducing the stale hand-maintained
//! parallel list this project keeps rediscovering. Assert rung identity and
//! ordering; never the number attached to a rung.
//!
//! **G1 is deliberately unguarded.** `binary_format.tex:3603` states
//! outright that genesis tranche G1 has no standalone Revision History row
//! — it is recorded retroactively *inside* the G2a row ("genesis tranche G1
//! --- landed at `3b09595` with no matching entry here"). Demanding a
//! principal marker for G1 would make this guard born red against a
//! document that pin B2 leaves correct, so only G2a, G-minor, G2b, G3a, and
//! G3b get principal-marker assertions here. Do not "fix" this by adding a
//! sixth marker; that rediscovers the contradiction pin B6 already resolved.
//!
//! **G3b** (`spec/CONTRACT_GENESIS_G3B_MEASURE.md` pin 14) extends this same
//! guard to its own Revision History row, closing the genesis ladder: the
//! contract's own M71 deletes the G3b row, observes this guard fail, and
//! restores it by hand.
//!
//! **P13-S21** extends the file past rung identity to one vocabulary gap of
//! the same shape: `PreconditionFailureReason` 14 and 15 entered at Push 4a
//! and reached this document at neither of the two places that owe them —
//! not the bounded enumeration that assigns discriminants, and not Push 4a's
//! own history row. See
//! `precondition_failure_reasons_14_and_15_are_documented` below.
use std::fs;
use std::path::Path;
/// The rung name immediately preceded by the row's `---` separator, in the
/// exact spelling each row uses (G-minor is never prefixed "Genesis
/// tranche" in the document; the other four are).
const PRINCIPAL_MARKERS: [(&str, &str); 5] = [
("G2a", "--- Genesis tranche G2a"),
("G-minor", "--- G-minor"),
("G2b", "--- Genesis tranche G2b"),
("G3a", "--- Genesis tranche G3a"),
("G3b", "--- Genesis tranche G3b"),
];
fn binary_format_source() -> String {
let path = Path::new(env!("CARGO_MANIFEST_DIR")).join("../../spec/binary_format.tex");
fs::read_to_string(&path).unwrap_or_else(|error| {
panic!("failed to read {}: {error}", path.display());
})
}
/// Collapse whitespace runs to a single space so a future rewrap of a row's
/// LaTeX source lines cannot silently break a substring search that this
/// guard depends on. Byte offsets after this pass are relative to the
/// normalized string, which is all the ordering assertion needs — nothing
/// here reports a source line number.
fn normalize_whitespace(text: &str) -> String {
let mut out = String::with_capacity(text.len());
let mut last_was_space = true;
for ch in text.chars() {
if ch.is_whitespace() {
if !last_was_space {
out.push(' ');
last_was_space = true;
}
} else {
out.push(ch);
last_was_space = false;
}
}
out
}
/// Every non-overlapping byte offset at which `needle` occurs in `haystack`.
fn find_all(haystack: &str, needle: &str) -> Vec<usize> {
let mut offsets = Vec::new();
let mut cursor = 0;
while let Some(relative) = haystack[cursor..].find(needle) {
let offset = cursor + relative;
offsets.push(offset);
cursor = offset + needle.len();
}
offsets
}
/// Slice the normalized document down to the Revision History chapter: from
/// its own `\chapter{Revision History}` heading to the next `\chapter{`
/// (there is none after it today, so this also tolerates a future chapter
/// being appended afterward) or the end of the document.
fn revision_history_slice(normalized: &str) -> &str {
const CHAPTER: &str = r"\chapter{Revision History}";
const NEXT_CHAPTER: &str = r"\chapter{";
let start = normalized
.find(CHAPTER)
.expect("binary_format.tex has no \\chapter{Revision History}");
let after = start + CHAPTER.len();
let end = normalized[after..]
.find(NEXT_CHAPTER)
.map(|relative| after + relative)
.unwrap_or(normalized.len());
&normalized[start..end]
}
/// Each rung's principal marker exactly once in the slice. A prose-only
/// mention (bare name, no preceding `---`) does not count, and neither
/// does a duplicated row.
#[test]
fn revision_history_has_exactly_one_principal_marker_per_rung() {
let source = binary_format_source();
let normalized = normalize_whitespace(&source);
let slice = revision_history_slice(&normalized);
for (rung, marker) in PRINCIPAL_MARKERS {
let offsets = find_all(slice, marker);
assert_eq!(
offsets.len(),
1,
"expected exactly one principal marker {marker:?} for rung {rung} in the \
Revision History chapter, found {} (offsets {offsets:?})",
offsets.len()
);
}
}
/// The five marked rungs appear in ladder order: G1 -> G2a -> G-minor ->
/// G2b -> G3a -> G3b. G1 has no marker of its own (see the module comment),
/// so this checks the remaining five.
#[test]
fn revision_history_rungs_are_strictly_ordered() {
let source = binary_format_source();
let normalized = normalize_whitespace(&source);
let slice = revision_history_slice(&normalized);
let offsets: Vec<(&str, usize)> = PRINCIPAL_MARKERS
.iter()
.map(|(rung, marker)| {
let found = find_all(slice, marker);
assert_eq!(
found.len(),
1,
"expected exactly one principal marker {marker:?} for rung {rung}, found \
{found:?}"
);
(*rung, found[0])
})
.collect();
for window in offsets.windows(2) {
let (earlier_rung, earlier_offset) = window[0];
let (later_rung, later_offset) = window[1];
assert!(
earlier_offset < later_offset,
"expected {earlier_rung} (offset {earlier_offset}) to precede {later_rung} \
(offset {later_offset}) in the Revision History chapter; ladder order is \
G2a < G-minor < G2b < G3a < G3b"
);
}
}
/// The G2b row states what G2b actually did (pin B3), and the search is
/// bounded to the G2b row's own segment — from its principal marker to the
/// next principal marker (or the slice end) — so that G3a's row, which also
/// names `OperationEnvelopeBlock` 3 and mentions G2b in prose, cannot
/// satisfy this after the G2b row itself is deleted. Unbounded searching is
/// exactly the hole P13-S17 was filed over.
#[test]
fn revision_history_g2b_row_states_what_g2b_did() {
let source = binary_format_source();
let normalized = normalize_whitespace(&source);
let slice = revision_history_slice(&normalized);
const G2B_MARKER: &str = "--- Genesis tranche G2b";
let g2b_offsets = find_all(slice, G2B_MARKER);
assert_eq!(
g2b_offsets.len(),
1,
"expected exactly one G2b principal marker, found {g2b_offsets:?}"
);
let g2b_start = g2b_offsets[0];
let after_g2b = g2b_start + G2B_MARKER.len();
let next_marker_offset = PRINCIPAL_MARKERS
.iter()
.filter_map(|(_, marker)| {
find_all(&slice[after_g2b..], marker)
.first()
.map(|relative| after_g2b + relative)
})
.min()
.unwrap_or(slice.len());
let row_segment = &slice[g2b_start..next_marker_offset];
assert!(
row_segment.contains("SetTuningContext"),
"G2b row segment does not name SetTuningContext: {row_segment:?}"
);
assert!(
row_segment.contains(r"\tablenums{34}"),
"G2b row segment does not carry discriminant 34: {row_segment:?}"
);
assert!(
row_segment.contains("TuningContextSettings"),
"G2b row segment does not name the TuningContextSettings payload subset: \
{row_segment:?}"
);
assert!(
row_segment.contains("OperationEnvelopeBlock")
&& row_segment.contains(r"\tablenums{2}~$\rightarrow$~\tablenums{3}"),
"G2b row segment does not record the OperationEnvelopeBlock accept-set raise \
2 -> 3: {row_segment:?}"
);
}
/// Slice one `longtable` row: from `marker` to the `\\` that terminates the
/// row. Rows in this document carry no internal `\\`, so the first
/// terminator after the marker ends that row and nothing else.
fn row_segment<'a>(haystack: &'a str, marker: &str) -> &'a str {
let offsets = find_all(haystack, marker);
assert_eq!(
offsets.len(),
1,
"expected exactly one occurrence of the row marker {marker:?}, found {} \
(offsets {offsets:?})",
offsets.len()
);
let start = offsets[0];
let end = haystack[start..]
.find(r"\\")
.map(|relative| start + relative)
.unwrap_or(haystack.len());
&haystack[start..end]
}
/// `PreconditionFailureReason` 14 (`AcousticRealizationPinned`) and 15
/// (`TranspositionOutOfRange`) entered the vocabulary at Push 4a — the
/// Operation Catalog documented both at its own 0.8.0, and `effect.rs` has
/// carried both ever since — but this document named neither anywhere, its
/// bounded enumeration running 13 straight to G3b's 16. That is P13-S21, and
/// it is why P13-S20's decoder could stop at 13 unchallenged: an implementer
/// reading only the wire specification would have built exactly that decoder
/// and been right.
///
/// **Both sites are checked, each bounded to its own row**, because either
/// alone is satisfiable by the wrong thing. The G3b history row and the
/// bounded enumeration both discuss `PreconditionFailureReason` at length, so
/// an unbounded search for these names would go green the moment any row
/// mentioned them — the same hole P13-S17 was filed over, one vocabulary
/// down. No document version number appears here either; the Push 4a row is
/// located by its version-free `--- Transpose algebra (Push~4a)` separator
/// form, for the reason the module comment gives.
///
/// The enumeration is asserted as a **discriminant/name adjacency**, since it
/// is the normative assignment and must be exact. The history row is asserted
/// by name and discriminant presence within its own bounded segment: it is
/// prose, its wording may legitimately be rewritten, and the bounding is what
/// makes presence load-bearing there.
#[test]
fn precondition_failure_reasons_14_and_15_are_documented() {
let source = binary_format_source();
let normalized = normalize_whitespace(&source);
let enumeration = row_segment(
&normalized,
r"\texttt{Precondition\allowbreak FailureReason} &",
);
for pairing in [
r"\tablenums{14} AcousticRealizationPinned",
r"\tablenums{15} TranspositionOutOfRange",
] {
assert!(
enumeration.contains(pairing),
"the bounded PreconditionFailureReason enumeration does not assign \
{pairing:?}; it is the normative discriminant assignment and the \
pairing must be exact: {enumeration:?}"
);
}
let push4a = row_segment(
revision_history_slice(&normalized),
"--- Transpose algebra (Push~4a)",
);
for owed in [
"AcousticRealizationPinned",
"TranspositionOutOfRange",
r"\tablenums{14}",
r"\tablenums{15}",
] {
assert!(
push4a.contains(owed),
"Push 4a's own Revision History row does not record {owed:?}. Both \
reasons were appended in Push 4a's epoch alongside OperationKind 30, \
and a row that records only the kind leaves the append unattributable: \
{push4a:?}"
);
}
}