What fails

import queue
b = queue.Queue(maxsize=2)      # app.py:575, verbatim

pascal26:2: error: Nil Python: Queue() is missing a value for parameter 1, which has no default

The shim declared exactly what the message asks for:

constructor Create; overload;
constructor Create(maxsize: Integer); overload;

queue.Queue(2) compiles and runs. The parameter is named maxsize. Renaming it from n to maxsize changed nothing, which is what sent this to a measurement rather than to a second guess.

The three controls, and what each rules out

Measured 2026-09-10, compiler 61f8a78f8aae, HEAD d326d8535.

probe	shape	result
`C(maxsize=5)` on `class C: def __init__(self, maxsize)`	NilPy-defined constructor	works
`q.put(v=7)` into a Pascal shim	Pascal METHOD, not a constructor	works
`array.array(tc="h")`	Pascal constructor, OVERLOADED — `Create(tc)` / `Create(tc, init)`	works
`queue.Queue(maxsize=2)`	Pascal constructor, overloaded — `Create` / `Create(maxsize)`	FAILS

So it is not keyword arguments in general, not Pascal callees in general, and not overloading in general — the third row is an overloaded Pascal constructor whose keyword binds correctly. The first draft of this ticket said "overloaded constructor" and that was wrong; mimic_array refutes it, and it is the row that stops this being filed against the wrong mechanism.

What separates the failing row from the passing one is that its overload set contains an arm taking no parameters at all, so parameter 1 of "the constructor" is not a single thing to bind a name against.

What was NOT measured, and a fix should establish it rather than assume this ticket did: whether a zero-parameter arm is precisely the trigger, or whether the real rule is something broader like "the arms disagree about the name of parameter 1". Two probes settle it — Create(a) / Create(b, c) with different first-parameter names, and Create / Create(a) where the keyword names a — and neither was run here because the workaround removed the need.

Why it is not the ticket next door

bug-n-an-overloaded-constructor-is-picked-by-name-ignoring-argument-type (prio 55) is about WHICH arm gets chosen when several could take the positional arguments given — first match by name, ignoring types. This one is about a keyword failing to bind after an arm is chosen; the diagnostic even names the chosen arm's parameter 1. They may share a cause in the same resolution path, and if a fix closes both that is a good outcome — but they fail differently and one is not evidence for the other.

Workaround in place

lib/rtl/mimic_queue.pas declares ONE constructor with a default:

constructor Create(maxsize: Integer = 0);

which serves Queue(), Queue(2) and Queue(maxsize=2) alike. Marked REVERT TO TWO OVERLOADS in the file, in the style of devdocs/dev/track-b-workarounds.md: it is a shape chosen to sidestep an open compiler bug, not the shape the code wants.

This is the second shim in two days to collapse a constructor overload set for a reason that is not about the shim — mimic_array did the same thing on 2026-09-09 for the sibling ticket. Two shims cannot both be wrong about how to write a constructor; the resolution path is.