[yeroo]

What and why

Opening a .mpp dropped task Notes, one of the task elements #417 made Task keep, so a .mpp opened and saved as .yppx/.xml lost them.

Current-layout .mpp files store a task's notes as RichEdit RTF under Var2Data key 0x000F (found by diffing corpus steps 06 → 07; the block is the RTF followed by \r\n and a NUL). This PR:

• decodes that key in taskdecode::var_fields into MppTaskFields.notes, and import_tasks copies it into Task.notes;
• adds mppread/src/rtf.rs, a dependency-free RTF → plain-text reader producing exactly the text Project's MSPDI export writes (final \par dropped, CRLF paragraphs). It hides non-text destinations (font/colour tables, pictures, objects, \bin payloads, bullet pntext, fields, headers…), honours \uc/\u, and decodes \'hh and raw 8-bit text through the current font's \cpg/\fcharset or \ansicpg;
• adds mppread/src/rtf_codepage.rs, the Windows single-byte code pages 874 and 1250–1258, generated from MultiByteToWideChar by corpus/tools/gen_rtf_codepages.py; Symbol-charset bytes map to U+F000+byte;
• refuses a note with "invalid notes for UID n" when its RTF is malformed, uses a DBCS/unknown code page for escaped bytes, or decodes to a character XML 1.0 cannot hold (consistent with how task names are validated);
• documents key 0x000F in corpus/mpp/README.md and adds Notes to PROJECT.md's list of fields that survive an import.

Not in this PR (no corpus case shows where Project stores them — tracked in a follow-up issue): Contact, WBSLevel, SubprojectName, PreLeveledStart/Finish, StatusManager, CommitmentStart/Finish, IsPublished (every snapshot row matches !summary && active, so no bit can be told apart from a summary/active marker without a Publish toggle case), non-default DisplayAsSummary/CommitmentType, and Notes on the legacy/MPP12 decode path.

Closes #455

Tests

• oracle_corpus: same!(notes) — every current-layout .mpp pair's decoded and imported Notes equal Project's MSPDI export (incl. step 07's First line.\r\nSecond line — unicode ✓ and «quotes».).
• taskdecode: the verbatim 314-byte corpus note decodes, imports, and survives write_mspdi → read_mspdi; corrupt RTF refuses with the UID.
• rtf.rs unit tests: RichEdit notes, bullets, pictures/objects/\bin, code pages and charsets (incl. brace-less font tables and nested \fontfile), \uc fallbacks, control symbols, malformed and XML-illegal input.
• Whole suite on the PR head: see the auto-merge note below. overalloc::tests::project_assignment_and_resource_offsets is skipped locally: with only the fetched snapshots/ corpus present it finds 47 < 50 Project pairs — pre-existing and unrelated (follow-up filed); CI has no corpus, where it returns early.

Review record

Plan agreed with Codex at v2 (v1 dropped IsPublished decoding for lack of a toggle case, and pinned the note block's \r\n\0 tail). Five revmux rounds:

• r1: 3 Major (picture/\bin payloads leaked or refused, \fcharset ignored, non-1252 locales refused the file), 4 Minor (fallback skipping, backslash-newline, test helper, docs) — all fixed.
• r2: 2 Major (bullet lists refused, \fcharset0 used \ansicpg), 4 Minor (\plain, unknown control symbols, Python vs Windows code-page tables, README escape), 3 Immaterial (\cpg, header check, duplicated lexer) — all fixed.
• r3: 1 Major (Thai cp874), 1 Minor (raw 8-bit text), 1 Immaterial (nested \fontfile\cpg) — all fixed.
• r4: 1 Minor (brace-less font tables, a regression from r3's fix) — fixed.
• r5: 2 Minor (nested \fontfile in a brace-less entry, XML-illegal characters) — fixed and verified by the planner's direct diff read.

No finding was disputed or deferred.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 93.36188% with 31 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
mppread/src/rtf.rs	93.85%	25 Missing ⚠️
mppread/src/rtf_codepage.rs	71.42%	6 Missing ⚠️

📢 Thoughts on this report? Let us know!

[yeroo]

Merged automatically (auto-merge and full autonomy are on for this checkout):

• Review clean after 5 revmux rounds. r1–r4 findings all fixed; the two r5 Minor findings were fixed in 5af37b7 and verified by the planner's direct diff read. None disputed, none deferred.
• Whole suite green on bda543b (the merge of main b43316e; one union conflict in mppread/src/lib.rs): 3976 tests, 0 failures, plus cargo fmt/clippy for the workspace and suite/, run exactly as CI runs them (without the private .mpp corpus). The corpus-backed oracle comparison, including the new Notes check, passed on 5af37b7 (3942 tests). It cannot run on bda543b because main's MPP import builds no assignment baselines (dates, work, cost, BCWS/BCWP) #445 refuses corpus 44-baseline.mpp independently of this PR — reproduced on a clean worktree of main and filed as bug: .mpp import refuses corpus 44-baseline.mpp since #445 (invalid baseline work for assignment UID 17) #523.
• CI: 11/11 checks passed on bda543b. merge-check: ok.

Follow-ups:

• bug: .mpp open still drops Contact, IsPublished and the other #417 task elements (need corpus cases) #521 — own issue: the other bug: saving drops a task's Notes and other unread task elements #417 task elements (Contact, WBSLevel, SubprojectName, PreLeveledStart/Finish, StatusManager, CommitmentStart/Finish, IsPublished, non-default DisplayAsSummary/CommitmentType) and Notes on the legacy/MPP12 path need corpus cases before they can be decoded (minor, from plan v2 out of scope).
• Leftovers from #455: bug: opening a .mpp drops task Notes and the other task elements #417 keeps #522 — Leftovers from #455: - [ ] the overalloc corpus test fails with only the fetched snapshots corpus (47 < 50 pairs), mppread/src/overalloc.rs:524 (minor).
• bug: .mpp import refuses corpus 44-baseline.mpp since #445 (invalid baseline work for assignment UID 17) #523 — main regression found during the UPDATE round (not from this PR).