Pascal corpus rung 1 — FPC test-suite subset (conformance)

• Type: feature (frontend conformance corpus)
• Track: P (Pascal frontend)
• Status: unfinished (parked) — rung-1 harness delivered and live (sweep 279/0, gap:/wontfix: taxonomy, twatch dashboard consumes conformance.tsv); remaining burn-down is rainy-day per user call 2026-07-11, tag: compat [[feature-pascal-corpus-expansion]].
• Owner: —

Idea

FPC ships tests/test/** — thousands of small .pp programs, each exercising one language feature (tbs*, tobject*, tgeneric*, terror* for expected failures, etc.). It is the c-testsuite analog for Pascal, but authoritative (the reference compiler's own suite) and far larger. PXX is FPC-faithful, so most should compile + run to the same result — and every one that doesn't is a sharply localized dialect bug.

Scope (start small, ladder up)

• Vendor a curated subset first (installer fetcher, pinned FPC release tag, gitignored) — not all thousands at once. Begin with the categories the compiler self-host never exercises: generics (tgeneric*), classes/properties/ visibility (tobject*, class*), exceptions (texcept*), RTL/string/ math units, operator overloading, variants.
• Runner (mirror install_lib_candidates.sh + the c-testsuite harness): compile each, run, diff stdout; honor the suite's %FAIL/%NORUN/expected-error markers so terror* negatives count as pass when they correctly reject.
• Skip list, burned ticket by ticket — same discipline as c-testsuite: a failing program → one narrowed frontend bug (Track P for lexer/parser/ dialect, Track A for IR/backend). Do NOT inline-fix during the audit; file, then resolve. Report the running pass count (like c-testsuite 220/220).

Watch-outs

• FPC test suite assumes FPC RTL + modes; gate each program on mode and unit availability — a test needing a unit PXX lacks is a skip (→ RTL/library ticket), not a frontend bug. Separate "dialect gap" from "missing RTL."
• Expected-failure tests (terror*) must be run for rejection, not acceptance — a PXX that accepts an invalid program is a real bug the suite catches.
• Keep the vendored tree gitignored; commit only the fetcher + the skip-list + pass-count report.

Gate

make test + self-host byte-identical for any frontend fix (shared lexer.inc/parser.inc). Rung is "green" at an agreed pass threshold on the curated subset; expand the subset as rungs clear.

Links

Rung of [[feature-pascal-corpus-expansion]] · method mirror [[feature-c-corpus-expansion]] · dialect policy [[project_fpc_compat_next_queue]] · [[project_mimic_fpc_done]].

2026-07-10 — infra landed + baseline audit

• Fetcher: tools/install_lib_candidates.sh fpc-testsuite — sparse tests/test + tests/tstunits (erroru et al, symlinked next to the tests) at FPC release_3_2_2 (0d122c49), gitignored, PROVENANCE.md.
• Runner: tools/run_pascal_conformance.sh — curated categories (generics, classes/props, exceptions, operators, strings/arrays/sets/case/enums/for-in, extended records, interfaces...), FPC dotest directives honored (%FAIL = must-reject, %NORUN, %RESULT, cpu/target/version/opt gates auto-skip), --shard I/N, --all, --only GLOB. Skip list test/pascal-conformance/pxx.skip (name + reason), same discipline as c-testsuite.
• Baseline: 222 pass / 294 skip-listed / 34 auto-gated of 550 curated.
• Cluster tickets filed: [[bug-pascal-headerless-program]] (111!), [[feature-pascal-delphi-generics-syntax]] (93), [[feature-pascal-generic-nonclass-templates]] (10), [[feature-pascal-class-management-operators]] (8), [[bug-pascal-missing-diagnostics-fail-tests]] (13), [[task-pascal-conformance-long-tail]] (rest).
• Note: the two big parser clusters touch shared parser.inc → sole-A confirmation needed before an E+B+P agent edits them.

Parked 2026-07-10

Infra + baseline + cluster tickets landed (see above). Burn-down of the two big parser clusters edits shared parser.inc — needs the sole-A confirmation an E+B+P agent doesn't have. Resume: grab a cluster ticket, confirm sole-A, burn skip-list entries.

Note 2026-07-11 — dialect-pass category + strict-case sweep

pxx.skip now distinguishes dialect-pass (PXX lax by design / FPC-impl-limit tests — NOT bugs, do not burn: tgeneric14/20/30, tgenconstraint38/39, tenum2, tover3) from accepts-invalid (real missing diagnostics, rainy-day). The sweep runs the compiler with --strict-case (FPC-parity case-label diagnostics; PXX's default keeps lax first-match labels). tclass13c (TRootClass.Integer nested-type member) is PARKED by user call — needs a per-class nested-type registry, near-zero value; revisit only if nested-type lookups matter elsewhere. Triage details in bug-pascal-missing-diagnostics-fail-tests.

2026-07-14 — UNPARKED, and the numbers in this ticket are stale.

The sole-A parking excuse is gone: both big parser clusters it was blocked on (bug-pascal-headerless-program, feature-pascal-delphi-generics-syntax) plus feature-pascal-generic-nonclass-templates are in done/.

Current sweep at HEAD: 283 pass / 7 fail / 226 skip (this ticket's text claims a 279/0 sweep, and the baseline it quotes is 222 — both stale). Today's session moved it 273 -> 283 by burning the whole advanced-record legality cluster (b347) plus tenum4 / the template cluster / tsealed (b342-b344).

The 7 that remain, correctly classified (the diagnostics umbrella had these filed as "13 of 17 = one visibility bug", which was wrong — only ONE of them is visibility):

• tclass12b — strict private const reached from a DESCENDANT. The only real visibility test. Access control is still not enforced anywhere.
• tclass14b — a published class property.
• toperator71 / 92 / 95 — operator declaration rules (a global operator = on a class type; two Implicit overloads differing only in RESULT type).
• tdefault8 — compile gap: a nested type reference (TTest.TRecord) -> unknown type.
• tset4 — compile gap: TSysCharSet is missing from the RTL.

The two compile gaps are the only ones that are not diagnostics, and they are the ones worth taking first — a program that will not compile is a bigger hole than a program we fail to reject.

Moved back to backlog/.

• 2026-07-19 (backlog sweep note) Stale numbers: sweep at HEAD is 323 pass / 0 fail (7b54288c) — the 283/7 and 279/0 figures above are outdated. Remaining scope = the ~200-entry skip-list burn-down (rainy-day per user).