Re-triage: 33 skipped conformance tests now pass by exit code

• Type: task (conformance bookkeeping — Track A/P follow-up of the 2026-07-14/15 night session). Per the skip-file header, unskipping is deliberate per-test review: the runner checks EXIT CODES and the %FAIL contract only, so an "exit 0" can still print wrong output (tforin24's skip reason documents exactly that trap — it is in this list).
• Opened: 2026-07-15 (night), from a skip-list-emptied sweep at 86c34639: curated 295 pass/0 fail; without the skip list 328 pass — 33 listed tests no longer fail.

Candidates (verify OUTPUT against FPC before unskipping each)

tarray1, tclass10c, tclass12a, terecs16, texception2, tforin14, tforin24 (KNOWN wrong-output trap), tforin25, tgeneric9, tgeneric28, tgeneric34, tgeneric35, tgeneric37, tgeneric38, tgeneric39, tgeneric40, tgeneric41, tgeneric43, tgeneric44, tgeneric45, tgeneric46, tgeneric47, tgeneric80, tgeneric81, tgeneric82, tgeneric89, tgeneric90, toperator12, toperator93, tprop, tsealed4, tstatic1, tstatic5

Likely green-lit by tonight's arcs: record value-typecast offsets, bitwise not family, qword domain/conversion fixes, {$Q+} + catchable EIntOverflow/EDivByZero, ValQWord + qword text I/O, on E do w/o var name, type-helper work from the prior session.

Method

For each: compile + run under pxx AND FPC, diff stdout byte-for-byte (the tint642 method); a {%FAIL} test must be REJECTED with the sweep's strict flags. Unskip only on a clean diff; else update the reason with the real residual. Expect several tgeneric* to be one shared fix — cluster.

Log

• 2026-07-15 — resolved, commit 7b54288c.