diff --git a/.github/workflows/nightly-property.yml b/.github/workflows/nightly-property.yml index 32ed103b..207972cc 100644 --- a/.github/workflows/nightly-property.yml +++ b/.github/workflows/nightly-property.yml @@ -2,14 +2,24 @@ name: Nightly property fuzz # The daily heavy property run for the ProseMirror<->Markdown converter # (packages/prosemirror-markdown). The PR/CI test run keeps NUM_RUNS modest to -# stay under budget; this cron cranks it up (~5000) with a RANDOM seed to hunt +# stay under budget; this cron cranks up total coverage with random seeds to hunt # for deeper round-trip counterexamples than a fixed-seed PR run can reach. # -# Counterexample -> fixture workflow: when a run fails, fast-check prints the -# SHRUNK minimal counterexample plus the reproducing seed. This job files a -# Gitea issue containing that seed + counterexample. A human then commits the -# shrunk doc as a PERMANENT fixture under -# packages/prosemirror-markdown/test/fixtures/counterexamples/ with a case in +# WHY SHARDING: a single mega-run (~10000 fast-check runs) OOMs the vitest worker +# (empirically ~1625 runs -> "JS heap out of memory", ~2GB) because heap +# accumulates across the whole property run in one process. Instead this job runs +# SHARDS fresh vitest processes, each a MODERATE per-shard count with a DISTINCT +# derived seed, so total coverage ~= SHARDS x PER_SHARD_NUM_RUNS across processes +# that never accumulate heap. On the first failing shard we stop and keep that +# shard's output for triage. +# +# Counterexample -> fixture workflow: when a shard fails, fast-check prints the +# SHRUNK minimal counterexample plus the reproducing seed. This job files a Gitea +# issue containing that seed + counterexample ONLY when the output actually holds +# a fast-check counterexample; an infra failure (OOM/tsc/install, no +# counterexample) is filed under a DISTINCT title so it can never poison the +# counterexample dedup. A human then commits the shrunk doc as a PERMANENT fixture +# under packages/prosemirror-markdown/test/fixtures/counterexamples/ with a case in # counterexamples.test.ts, and FIXES the converter (never weakens a property to # hide the bug). See packages/prosemirror-markdown/README.md. @@ -20,11 +30,11 @@ on: workflow_dispatch: inputs: num_runs: - description: 'fast-check runs per property' + description: 'fast-check runs PER SHARD (8 shards run in sequence)' required: false - default: '5000' + default: '600' seed: - description: 'fast-check seed (empty = random)' + description: 'base fast-check seed (empty = random); shard i uses base+i' required: false default: '' @@ -57,60 +67,152 @@ jobs: # against source without the package's build/. Skipping the build also # keeps a tsc build error from masquerading as a property-test failure and # filing a bogus counterexample issue. - - name: Resolve seed and run count + - name: Resolve base seed and per-shard run count id: params + # Dispatch inputs are read via env (NOT interpolated into the shell body) + # to avoid script injection through a crafted input value. + env: + SEED_INPUT: ${{ inputs.seed }} + NUM_RUNS_INPUT: ${{ inputs.num_runs }} run: | set -euo pipefail - SEED="${{ inputs.seed }}" + SEED="${SEED_INPUT:-}" # Empty seed (cron, or a dispatch that left it blank) -> random. Combine # two RANDOMs so the seed spans more than RANDOM's 0..32767 range. [ -z "$SEED" ] && SEED=$(( (RANDOM << 15) | RANDOM )) - NUM_RUNS="${{ inputs.num_runs || '5000' }}" + NUM_RUNS="${NUM_RUNS_INPUT:-}" + [ -z "$NUM_RUNS" ] && NUM_RUNS=600 echo "seed=$SEED" >> "$GITHUB_OUTPUT" echo "num_runs=$NUM_RUNS" >> "$GITHUB_OUTPUT" - echo "Property fuzz: PROPERTY_SEED=$SEED PROPERTY_NUM_RUNS=$NUM_RUNS" + echo "Sharded property fuzz: BASE_SEED=$SEED PER_SHARD_NUM_RUNS=$NUM_RUNS SHARDS=8" - - name: Run generative property suite + - name: Run generative property suite (sharded) + id: fuzz env: - PROPERTY_SEED: ${{ steps.params.outputs.seed }} - PROPERTY_NUM_RUNS: ${{ steps.params.outputs.num_runs }} + BASE_SEED: ${{ steps.params.outputs.seed }} + PER_SHARD_NUM_RUNS: ${{ steps.params.outputs.num_runs }} + SHARDS: '8' run: | - set -o pipefail - pnpm --filter @docmost/prosemirror-markdown exec \ - vitest run test/generative/ 2>&1 | tee property-output.txt + set -uo pipefail + # Give each fresh process headroom, but rely on SHARDING (not a big heap) + # to avoid OOM: a moderate per-shard count in a process that starts clean. + export NODE_OPTIONS=--max-old-space-size=4096 + : > property-output.txt + FAILED=0 + FAIL_SEED="" + i=0 + while [ "$i" -lt "$SHARDS" ]; do + SHARD_SEED=$(( BASE_SEED + i )) + echo "=== shard $((i + 1))/$SHARDS: PROPERTY_SEED=$SHARD_SEED PROPERTY_NUM_RUNS=$PER_SHARD_NUM_RUNS ===" + # tee OVERWRITES property-output.txt each shard; since we break on the + # first failure, the file ends up holding exactly the failing shard's + # output (which carries the shrunk counterexample + reproducing seed). + if PROPERTY_SEED="$SHARD_SEED" PROPERTY_NUM_RUNS="$PER_SHARD_NUM_RUNS" \ + pnpm --filter @docmost/prosemirror-markdown exec \ + vitest run test/generative/ 2>&1 | tee property-output.txt; then + echo "shard $((i + 1)) passed" + else + echo "shard $((i + 1)) FAILED (seed=$SHARD_SEED) — stopping; keeping its output" + FAILED=1 + FAIL_SEED="$SHARD_SEED" + break + fi + i=$(( i + 1 )) + done + echo "failed=$FAILED" >> "$GITHUB_OUTPUT" + echo "fail_seed=$FAIL_SEED" >> "$GITHUB_OUTPUT" + exit "$FAILED" - # On failure fast-check has already printed the shrunk counterexample and - # its reproducing seed into property-output.txt. File a Gitea issue with - # the seed, run count and the tail of that output so a human can turn the - # counterexample into a permanent fixture (see the header comment). - - name: File issue on failure - if: failure() + # A GENUINE counterexample: fast-check printed a shrunk minimal case and its + # reproducing seed into property-output.txt. File a dedup-guarded issue whose + # title prefix is UNIQUE to counterexamples, so an infra failure (handled by + # the next step under a different title) can never poison this dedup. + - name: File counterexample issue + # always() is REQUIRED: the fuzz step exits nonzero on a failing shard, + # so a bare `if:` (implicitly success() && ...) would skip this step + # exactly when it must run. always() lets it run on the failure path. + if: always() && steps.fuzz.outputs.failed == '1' env: - SEED: ${{ steps.params.outputs.seed }} + FAIL_SEED: ${{ steps.fuzz.outputs.fail_seed }} NUM_RUNS: ${{ steps.params.outputs.num_runs }} RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }} GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} + TITLE_PREFIX: 'Nightly property counterexample' run: | set -uo pipefail - TITLE="Nightly property test failure (seed=${SEED})" + # Discriminate counterexample vs infra failure by the fast-check + # signature. No signature -> leave it to the infra-failure step. + if ! grep -Eq 'Property failed after|Counterexample' property-output.txt; then + echo "No fast-check counterexample signature — infra failure, handled by the next step." + exit 0 + fi + TITLE="${TITLE_PREFIX} (seed=${FAIL_SEED})" - # Best-effort dedup: skip if an open issue with the shared title prefix - # already exists. A failure of this check must NOT block issue creation. + # Best-effort dedup: skip if an open issue with the counterexample title + # prefix already exists. A failure of this check must NOT block creation. EXISTING="" if EXISTING=$(curl -sS \ -H "Authorization: token ${GITHUB_TOKEN}" \ "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/issues?state=open&limit=100"); then if printf '%s' "$EXISTING" \ - | node -e 'let s="";process.stdin.on("data",d=>s+=d).on("end",()=>{let a;try{a=JSON.parse(s)}catch{process.exit(1)}if(!Array.isArray(a))process.exit(1);process.exit(a.some(i=>typeof i.title==="string"&&i.title.startsWith("Nightly property test failure"))?0:1)})'; then - echo "An open 'Nightly property test failure' issue already exists — skipping creation." + | node -e 'let s="";process.stdin.on("data",d=>s+=d).on("end",()=>{let a;try{a=JSON.parse(s)}catch{process.exit(1)}if(!Array.isArray(a))process.exit(1);const p=process.env.TITLE_PREFIX;process.exit(a.some(i=>typeof i.title==="string"&&i.title.startsWith(p))?0:1)})'; then + echo "An open '${TITLE_PREFIX}' issue already exists — skipping creation." exit 0 fi fi # Build the JSON body with the test output SAFELY escaped (never hand- # interpolate the counterexample into JSON). - BODY_TEXT=$(printf 'The nightly property fuzz run failed.\n\n- seed: `%s`\n- NUM_RUNS: `%s`\n- run: %s\n\nReproduce locally:\n\n```\nPROPERTY_SEED=%s PROPERTY_NUM_RUNS=%s pnpm --filter @docmost/prosemirror-markdown exec vitest run test/generative/\n```\n\nfast-check shrinks the failure to a minimal counterexample. Commit it as a permanent fixture under `packages/prosemirror-markdown/test/fixtures/counterexamples/` + a case in `counterexamples.test.ts`, then FIX the converter (do not weaken a property). See `packages/prosemirror-markdown/README.md`.\n\nTail of the test output (contains the shrunk counterexample):\n\n```\n%s\n```\n' \ - "$SEED" "$NUM_RUNS" "$RUN_URL" "$SEED" "$NUM_RUNS" "$(tail -n 120 property-output.txt)") + BODY_TEXT=$(printf 'A nightly property fuzz SHARD failed with a fast-check counterexample.\n\n- failing shard seed: `%s`\n- NUM_RUNS (per shard): `%s`\n- run: %s\n\nReproduce locally:\n\n```\nPROPERTY_SEED=%s PROPERTY_NUM_RUNS=%s pnpm --filter @docmost/prosemirror-markdown exec vitest run test/generative/\n```\n\nfast-check shrinks the failure to a minimal counterexample. Commit it as a permanent fixture under `packages/prosemirror-markdown/test/fixtures/counterexamples/` + a case in `counterexamples.test.ts`, then FIX the converter (do not weaken a property). See `packages/prosemirror-markdown/README.md`.\n\nTail of the test output (contains the shrunk counterexample):\n\n```\n%s\n```\n' \ + "$FAIL_SEED" "$NUM_RUNS" "$RUN_URL" "$FAIL_SEED" "$NUM_RUNS" "$(tail -n 120 property-output.txt)") + + jq -n --arg title "$TITLE" --arg body "$BODY_TEXT" \ + '{title: $title, body: $body}' > payload.json + + curl -sS -X POST \ + "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/issues" \ + -H "Authorization: token ${GITHUB_TOKEN}" \ + -H 'Content-Type: application/json' \ + -d @payload.json + + # An INFRA failure (OOM, tsc, install) has NO counterexample signature. File + # it under a DISTINCT title so it is visible but keeps the counterexample + # dedup (above) uncontaminated — a real counterexample can still file even + # while an infra issue is open. + - name: File infra failure issue + # always() is REQUIRED: the fuzz step exits nonzero on a failing shard, + # so a bare `if:` (implicitly success() && ...) would skip this step + # exactly when it must run. always() lets it run on the failure path. + if: always() && steps.fuzz.outputs.failed == '1' + env: + FAIL_SEED: ${{ steps.fuzz.outputs.fail_seed }} + NUM_RUNS: ${{ steps.params.outputs.num_runs }} + RUN_URL: ${{ github.server_url }}/${{ github.repository }}/actions/runs/${{ github.run_id }} + GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }} + TITLE_PREFIX: 'Nightly property run infra failure' + run: | + set -uo pipefail + # Only file when there is NO counterexample signature (else the + # counterexample step owns it). + if grep -Eq 'Property failed after|Counterexample' property-output.txt; then + echo "Counterexample present — owned by the counterexample step." + exit 0 + fi + TITLE="${TITLE_PREFIX} (seed=${FAIL_SEED})" + + EXISTING="" + if EXISTING=$(curl -sS \ + -H "Authorization: token ${GITHUB_TOKEN}" \ + "${GITHUB_API_URL}/repos/${GITHUB_REPOSITORY}/issues?state=open&limit=100"); then + if printf '%s' "$EXISTING" \ + | node -e 'let s="";process.stdin.on("data",d=>s+=d).on("end",()=>{let a;try{a=JSON.parse(s)}catch{process.exit(1)}if(!Array.isArray(a))process.exit(1);const p=process.env.TITLE_PREFIX;process.exit(a.some(i=>typeof i.title==="string"&&i.title.startsWith(p))?0:1)})'; then + echo "An open '${TITLE_PREFIX}' issue already exists — skipping creation." + exit 0 + fi + fi + + BODY_TEXT=$(printf 'A nightly property fuzz SHARD failed WITHOUT a fast-check counterexample (infra failure: OOM / build / install). This is NOT a converter round-trip bug.\n\n- failing shard seed: `%s`\n- NUM_RUNS (per shard): `%s`\n- run: %s\n\nInvestigate the run log (memory, dependency install, or a tsc/import error). The nightly counterexample dedup is intentionally separate from this issue.\n\nTail of the test output:\n\n```\n%s\n```\n' \ + "$FAIL_SEED" "$NUM_RUNS" "$RUN_URL" "$(tail -n 120 property-output.txt)") jq -n --arg title "$TITLE" --arg body "$BODY_TEXT" \ '{title: $title, body: $body}' > payload.json diff --git a/packages/prosemirror-markdown/src/lib/markdown-converter.ts b/packages/prosemirror-markdown/src/lib/markdown-converter.ts index 27363617..8389d2b0 100644 --- a/packages/prosemirror-markdown/src/lib/markdown-converter.ts +++ b/packages/prosemirror-markdown/src/lib/markdown-converter.ts @@ -640,9 +640,12 @@ export function convertProseMirrorToMarkdown( // ("5." starts the list at 5), and marked parses that back into //
    . Emitting `${start + index}.` keeps a start=5 list as // "5.","6.",… while a default (start=1) list stays "1.","2.",… See #351. - // `Number(...) || 1` coerces a numeric-or-absent start and guards against - // a stray non-numeric `start` string-concatenating into the marker. - const start = Number(node.attrs?.start) || 1; + // Only an INTEGER ≥ 2 is a valid explicit start; collapse everything else + // (absent, non-number, 0, negative, fractional) to 1. A negative/fractional + // start would otherwise emit an untokenizable marker (e.g. "-3.", "2.5.") + // that marked cannot parse, re-importing the whole list as a PARAGRAPH. + const raw = node.attrs?.start; + const start = Number.isInteger(raw) && (raw as number) > 1 ? (raw as number) : 1; return nodeContent .map((item: any, index: number) => processListItem(item, `${start + index}.`), @@ -1413,7 +1416,11 @@ export function convertProseMirrorToMarkdown( // Carry a non-1 `start` on the raw-HTML path (columns/spanned cells) via // the
      attribute, which the tiptap parser reads back into // attrs.start. A default (start=1) list stays a bare
        . See #351. - const start = block.attrs?.start ?? 1; + // Use the SAME integer-≥-2 guard as the markdown path: a fractional start + // would emit `start="2.5"` → parseInt→2 on import (path divergence), and a + // 0/negative start is not a valid explicit start either. + const raw = block.attrs?.start; + const start = Number.isInteger(raw) && (raw as number) > 1 ? (raw as number) : 1; const startAttr = start > 1 ? ` start="${start}"` : ""; return `${children .map((li: any) => `
      1. ${blockChildrenToHtml(li)}
        2. `) diff --git a/packages/prosemirror-markdown/test/generative/attr-arbitraries.ts b/packages/prosemirror-markdown/test/generative/attr-arbitraries.ts index c3c9a03d..f5cd05e3 100644 --- a/packages/prosemirror-markdown/test/generative/attr-arbitraries.ts +++ b/packages/prosemirror-markdown/test/generative/attr-arbitraries.ts @@ -37,14 +37,19 @@ * tagged `// ACCEPTED:` inline. Freezing them is correct — there is * nothing to preserve in the target format. * - * (b) PINNED BUG — the attribute IS representable in markdown but the - * converter drops it anyway (a real defect). These are NOT silently - * frozen: each is captured as a LOUD `it.fails` counterexample in - * test/fixtures/counterexamples/ + counterexamples.test.ts, and the - * freeze here only keeps the P1/P2 union green until a MAINTAINER rules - * on accept-vs-fix (the epic guardrail reserves that call). These: - * `column.width` (parseFloat drops `%`), `orderedList.start` (non-1 - * start renders as `1.`). Tagged `// PINNED-BUG:` inline. + * (b) FIXED & VALUE-FUZZED — attributes that were once PINNED converter + * bugs (representable in markdown but dropped) and are now FIXED in + * src/, so they are value-fuzzed here at legal non-default values like + * any healthy attr. `orderedList.start` (the non-1 start once rendered + * as `1.`; the converter now emits the start marker / `
          `) + * and `column.width` (a unitless flex-grow number that round-trips via + * parseFloat) are both fuzzed in OVERRIDES below. The former held-out + * `it.fails` cases are gone; `ordered-list-start.json` + + * counterexamples.test.ts now stand as PASSING regression pins (per the + * epic guardrail, the minimal doc stays forever to guard re-regression). + * The #351 media-family sizing attrs (image/video/youtube/pdf/drawio/ + * excalidraw/embed width/height/size/aspectRatio) are likewise fuzzed + * now that they ride round-trip-safely in the per-node `` JSON. * * (c) DEFERRED-BUG — representable AND round-trips, frozen only because the * flat generator can't yet build a valid instance. Table @@ -64,8 +69,10 @@ * number re-parses as a string), EXCEPT `embed.width/height` which the * embed schema keeps numeric — handled per-attr. * - * Both PINNED-BUG attrs (`column.width` P2 churn, `orderedList.start` P1 loss) - * are captured as committed `it.fails` counterexamples — NOT hidden here. + * The two former PINNED-BUG attrs (`column.width` P2 churn, `orderedList.start` + * P1 loss) are now FIXED and value-fuzzed; `ordered-list-start.json` in + * counterexamples.test.ts is a permanent PASSING regression pin, not an + * `it.fails` hold-out. */ import fc from 'fast-check'; import { getSchema } from '@tiptap/core'; @@ -142,7 +149,7 @@ const OVERRIDES: Record = { // FIXED (#351): the converter now emits the start marker ("5." /
            ) // and it round-trips, so the start number is value-fuzzed. See // counterexamples.test.ts (ordered-list-start.json) for the regression pin. - 'orderedList.start': { always: true, arb: num(2, 3, 5) }, + 'orderedList.start': { always: true, arb: num(2, 3, 5, 42) }, 'orderedList.type': { frozen: true }, // ACCEPTED: a/A/i markers not expressible in GFM 'taskItem.checked': { always: true, arb: fc.constant(true) }, // boolean, default false // ── codeBlock ──────────────────────────────────────────────────────────── diff --git a/packages/prosemirror-markdown/test/generative/env-int.test.ts b/packages/prosemirror-markdown/test/generative/env-int.test.ts new file mode 100644 index 00000000..d23a45d9 --- /dev/null +++ b/packages/prosemirror-markdown/test/generative/env-int.test.ts @@ -0,0 +1,29 @@ +import { describe, expect, it } from 'vitest'; +import { envInt } from './env-int.js'; + +// Unit coverage for the shared PROPERTY_SEED / PROPERTY_NUM_RUNS parser. The key +// contract is that an explicit "0" is honored (a valid fast-check seed) while +// empty/absent/non-numeric fall back to the default. +describe('envInt', () => { + it('honors an explicit "0" (does not fall back)', () => { + expect(envInt('0', 42)).toBe(0); + }); + it('falls back on empty string', () => { + expect(envInt('', 42)).toBe(42); + }); + it('falls back on undefined', () => { + expect(envInt(undefined, 42)).toBe(42); + }); + it('falls back on a non-numeric string', () => { + expect(envInt('abc', 42)).toBe(42); + }); + it('parses a plain integer string', () => { + expect(envInt('300', 42)).toBe(300); + }); + it('parses a negative integer', () => { + expect(envInt('-5', 42)).toBe(-5); + }); + it('parses a large integer', () => { + expect(envInt('1073741824', 42)).toBe(1073741824); + }); +}); diff --git a/packages/prosemirror-markdown/test/generative/env-int.ts b/packages/prosemirror-markdown/test/generative/env-int.ts new file mode 100644 index 00000000..9c7df021 --- /dev/null +++ b/packages/prosemirror-markdown/test/generative/env-int.ts @@ -0,0 +1,10 @@ +/** + * Parse an integer-ish environment variable with a default fallback. + * + * Used to read PROPERTY_SEED / PROPERTY_NUM_RUNS in the generative property + * suites. An unset/empty/non-numeric value falls back to `dflt`, but an explicit + * `"0"` is honored (a valid fast-check seed) — `Number(x) || dflt` would wrongly + * swallow 0. See flat-roundtrip.property.test.ts / nested-roundtrip.property.test.ts. + */ +export const envInt = (v: string | undefined, dflt: number): number => + v !== undefined && v !== '' && Number.isFinite(Number(v)) ? Number(v) : dflt; diff --git a/packages/prosemirror-markdown/test/generative/flat-roundtrip.property.test.ts b/packages/prosemirror-markdown/test/generative/flat-roundtrip.property.test.ts index 08aa7ac2..87c6093a 100644 --- a/packages/prosemirror-markdown/test/generative/flat-roundtrip.property.test.ts +++ b/packages/prosemirror-markdown/test/generative/flat-roundtrip.property.test.ts @@ -18,6 +18,7 @@ import { coveredTypes, KNOWN_UNCOVERED, } from './node-generators.js'; +import { envInt } from './env-int.js'; // ── Attribute-value coverage allowlist ────────────────────────────────────── // The node/mark completeness contract guarantees every TYPE is generated, but @@ -122,8 +123,7 @@ vi.setConfig({ testTimeout: 30000 }); // OOMs the worker). // An unset/empty/non-numeric value falls back to the default; an explicit 0 is // honored (a valid fast-check seed) — `Number(x) || default` would swallow it. -const envInt = (v: string | undefined, dflt: number): number => - v !== undefined && v !== '' && Number.isFinite(Number(v)) ? Number(v) : dflt; +// The parser is shared with the nested suite (env-int.ts) and unit-tested there. const SEED = envInt(process.env.PROPERTY_SEED, 20250705); const NUM_RUNS = envInt(process.env.PROPERTY_NUM_RUNS, 300); diff --git a/packages/prosemirror-markdown/test/generative/nested-roundtrip.property.test.ts b/packages/prosemirror-markdown/test/generative/nested-roundtrip.property.test.ts index d22dbf77..2a8a6768 100644 --- a/packages/prosemirror-markdown/test/generative/nested-roundtrip.property.test.ts +++ b/packages/prosemirror-markdown/test/generative/nested-roundtrip.property.test.ts @@ -11,6 +11,7 @@ import { } from '../../src/lib/index.js'; import { firstDivergence } from '../roundtrip-helpers.js'; import { schema, docArb } from './doc-generator.js'; +import { envInt } from './env-int.js'; // Each run does a real convert + jsdom parse; give ample headroom so the suite // is deterministic under parallel worker load (matching the flat sibling suite). @@ -39,8 +40,7 @@ vi.setConfig({ testTimeout: 60000 }); // random seed to hunt for deeper counterexamples. // An unset/empty/non-numeric value falls back to the default; an explicit 0 is // honored (a valid fast-check seed) — `Number(x) || default` would swallow it. -const envInt = (v: string | undefined, dflt: number): number => - v !== undefined && v !== '' && Number.isFinite(Number(v)) ? Number(v) : dflt; +// The parser is shared with the flat suite (env-int.ts) and unit-tested there. const SEED = envInt(process.env.PROPERTY_SEED, 20250705); // The nested walk builds far heavier docs than the flat suite (each P1/P2 run // parses the emitted markdown through jsdom), so keep the run count moderate to diff --git a/packages/prosemirror-markdown/test/ordered-list-start-normalization.test.ts b/packages/prosemirror-markdown/test/ordered-list-start-normalization.test.ts new file mode 100644 index 00000000..03593afc --- /dev/null +++ b/packages/prosemirror-markdown/test/ordered-list-start-normalization.test.ts @@ -0,0 +1,75 @@ +import { describe, expect, it } from 'vitest'; +import { convertProseMirrorToMarkdown } from '../src/lib/markdown-converter.js'; +import { markdownToProseMirror } from '../src/lib/markdown-to-prosemirror.js'; + +// --------------------------------------------------------------------------- +// #351 regression — DEGENERATE orderedList `start` normalization (DO-1). +// +// Only an INTEGER >= 2 is a valid explicit start. A degenerate start (0, +// negative, fractional, non-number) MUST collapse to the default 1 on export: +// emitting the raw value would produce an untokenizable marker (e.g. "-3.", +// "2.5.") that `marked` cannot parse, causing the WHOLE orderedList (and its +// listItems) to re-import as a PARAGRAPH — structural corruption. This test +// pins that each degenerate start: +// (1) exports to "1."/"2." markers (default numbering), +// (2) re-imports as a VALID `orderedList` (not a paragraph) with `start` +// defaulting, structure preserved, +// (3) is byte-stable (md1 === md2). +// +// Mutation check: revert the converter guard to `Number(raw) || 1` and this +// test goes RED (a fractional/negative start leaks into the marker). +// --------------------------------------------------------------------------- + +const makeDoc = (start: unknown) => ({ + type: 'doc', + content: [ + { + type: 'orderedList', + attrs: { type: null, start }, + content: [ + { + type: 'listItem', + content: [ + { type: 'paragraph', content: [{ type: 'text', text: 'alpha' }] }, + ], + }, + { + type: 'listItem', + content: [ + { type: 'paragraph', content: [{ type: 'text', text: 'beta' }] }, + ], + }, + ], + }, + ], +}); + +describe('#351 orderedList degenerate start normalization', () => { + for (const start of [0, -3, 2.5]) { + it(`start=${start} exports as default "1."/"2." markers`, () => { + const md = convertProseMirrorToMarkdown(makeDoc(start)); + expect(md).toContain('1. alpha'); + expect(md).toContain('2. beta'); + // The degenerate value must NOT leak into any marker. + expect(md).not.toMatch(/-?\d*\.\d+\.\s/); // no fractional marker like "2.5." + expect(md).not.toContain('-3.'); + }); + + it(`start=${start} re-imports as a valid orderedList (not a paragraph)`, async () => { + const md = convertProseMirrorToMarkdown(makeDoc(start)); + const doc = (await markdownToProseMirror(md)) as any; + const top = doc.content?.[0]; + expect(top?.type).toBe('orderedList'); + // start defaults (never the degenerate value); tiptap materializes 1. + expect(top?.attrs?.start ?? 1).toBe(1); + expect(top?.content?.length).toBe(2); + }); + + it(`start=${start} is byte-stable across a re-export`, async () => { + const md1 = convertProseMirrorToMarkdown(makeDoc(start)); + const doc = await markdownToProseMirror(md1); + const md2 = convertProseMirrorToMarkdown(doc); + expect(md2).toBe(md1); + }); + } +});