← board

Track guessed as P from the test source. The ranker reads frontmatter, so an unset track parks a stub in Track T's queue regardless of what the body says -- correct the track: line if this is wrong.

origin/master has advanced 1 commit(s) since this sha. Re-verify at current HEAD before acting — the callback is tagged to the sha that was tested, which may no longer be the state of the tree.

regression: test-core#src:test/test_pascal_cast_chain_method_call.pas red at 97f96a5cc766 (auto-filed by twatch)

Repro

tools/testmgr.py --tier native --job 'test-core#src:test/test_pascal_cast_chain_method_call.pas' at 97f96a5cc7664fb2521b3fa5d0bd5de681b6c0d2

Range

The named sha 97f96a5cc766 CANNOT be the cause — it touches no buildable file (docs / tickets / tstate only). It is the sha that was TESTED, i.e. the upper bound of an untested range; the cause is somewhere below it.

bad 97f96a5cc766, last good 67a83ca0bb63, 2 commit(s) in range — the watcher narrows this by idle bisect; check tstate/TSTATE.md for the current range.

Log tail

pascal26:100: error: "NewN": this value has no members (only records, classes, interfaces and variants do)
pascal26:100: error: expected comma or close parenthesis
(tail)
pascal26:100: error: "NewN": this value has no members (only records, classes, interfaces and variants do)
  near: q    cr  >>> NewN   
pascal26:100: error: expected comma or close parenthesis
  near:    cr  NewN >>>

Stub ticket: signal only. Track T agent (face 2) enriches or a dev track takes it from the repro line.

Triaged and FIXED, 2026-08-27

Cause: acd14fe64 (fix(parser): route record-cast postfix walkers through ParseClassRecordSelectors), which is in the watcher's 2-commit range. Confirmed by reading the change rather than by bisect: that commit removed the FindUField(...) < 0 guard on the record-cast rvalue arm so every .name delegates to ParseClassRecordSelectors. Removing it was correct — the guard is what made TLongWordRec(l).b[0] read a Byte at the record's width and return 18486690310128388 — but it changed which walker sees a chain that starts with a field.

Repro, reduced

type
  TBaseClass = class of TBase;
  TRec = record cr: TBaseClass; n: Integer; end;
...
  writeln(TRec(q^).cr.NewN(3));   { pascal26: "NewN": this value has no members }

p^.cr.NewN(3) and PRec(q)^.cr.NewN(3) were unaffected — they take other walkers. Only the record-cast spelling changed hands.

A class of T field records REC_NONE (the class lives in UFldPtrElemRec), so after the .cr step recId is REC_NONE, ci stays -1, and every arm in ParseClassRecordSelectors declined. The arm it needed did not exist there.

Fix

compiler/pasparser_lval.inc, ParseClassRecordSelectors — a metaclass receiver arm, placed before the ci computation so it sees the node first.

The three other walkers that can reach a metaclass value each already have this arm, and all three go through the same shared pair: NodeMetaclassCi (the five-spelling predicate) and ParseMetaclassMemberTail (the member parser). This is the fourth caller of that pair, not a fourth mechanism — which is the shape devdocs/dev/root-cause-over-microfix.md asks for, and the reason the alternative (restoring the field guard) was not taken: it would have brought the byte-width read back with it.

Verified