← board

An unranked terminal folder is the wrong home for a measurement that explains a live red

The fork

The measurement is TRUE at 0bbd82cd7 (compiler/pascal26 sha 7fca108e4b85):

unit g1; {$mode objfpc} interface
generic procedure Test<T>(a: T);
implementation
generic procedure Test<S>(a: S); begin WriteLn('ran'); end;
end.

specialize Test<Integer>(1) from an importer prints ran. FPC rejects it.

The scope rule says reject. CLAUDE.md: "Us accepting what FPC rejects is not a defect." The goal file adds the sharper test — an input "whose two readings differ only when the program is already wrong." A renamed type parameter is produced by a mistake and by nothing else; the positional binding pxx uses gives that source the meaning its author intended. There is no real source that WANTS the rename diagnosed, and "evidence that settles it is real source that wants the behaviour."

What stops the rejection is the folder, not the reasoning. rejected/ is loaded so citations resolve, and it is not ranked. Two conformance rows — tgenfunc17.pp and tgenfunc18.pp — fail on this and will keep failing. A reader of that run finds two reds, greps the backlog, finds nothing ranked, and files the ticket again. The rejection is correct and it destroys the pointer.

Options

(a) known-incompat/, with the two conformance rows cited as expected FAILs. That folder is for a measurement that is true and reproducible and still not a defect, recorded as CHOSEN rather than tolerated — which is exactly this. The two reds become documented expectations with a citation, and the run's shape (347/2) is explained rather than merely observed.

(b) Implement the check. Cheap, zero runtime cost, and defensible on grounds that have nothing to do with FPC parity: a diagnostic that catches a certain mistake is worth having on its own. The cost is that it concedes the parity framing the goal file is trying to retire.

Recommendation: (a). It is what known-incompat/ is for, and it is the only option that fixes the residual question — why are these two rows red — rather than only the ticket. (b) is not wrong, but it settles a specific case while leaving the general one open.

The general question, which is the reason this is in U and not just a ticket move

A terminal folder that is loaded but unranked is the right home for a WRONG report and the wrong home for a CORRECT measurement that explains a live red.

rejected/ currently carries both meanings. The four terminal folders are documented as saying different things, and they do — but the axis they are cut on is is the report right, while the axis that matters for a red test row is does anything still point at this. A correct-but-chosen divergence with a failing conformance row attached needs to stay findable from the red, and known-incompat/ is the only unranked folder whose name carries that claim.

If U agrees, the durable output is not this ticket's placement but a line in the folder rules: before filing to rejected/, ask whether any red test row cites the behaviour. If one does, the folder is known-incompat/.