[bug] t27o: the int_type_names splice refusal tests the mod alias, not the #[path] module, so mod names; slips through #10843

What is wrong

path (specs/tools/t27o/rust_convert.t27) refuses a call into the generated module int_type_names (R_SPLICE: the module's own use type_text:: stays unresolved when t27c splices it, so the spec test-report is BLOCKED -- the gap specs/compiler/int_type_names.t27 records). It decides that with

if (word_is(src, st[s], st[s + 1], "int_type_names")) { refuse(st, R_SPLICE, a, a, mb); return; }

but st[s] .. st[s + 1] of a K_MODPATH entry is the alias the mod item declares (ma .. mb), not the module: the module is the #[path] string, held in the owner slots (st[s + 2] .. st[s + 3], e.g. compiler/int_type_names). The check therefore passes whenever the alias is not literally int_type_names.

Repro (t27o of the branch of #10817)

#[path = "../../../gen/rust/compiler/int_type_names.rs"]
mod names;
fn suffix(t: &str) -> bool { names::is_integer_type_suffix(t.as_bytes()) }

tri t27o converts suffix and emits use compiler::int_type_names; plus int_type_names::is_integer_type_suffix(t) -- exactly the spec the refusal exists to prevent. With mod int_type_names; the same fn is refused with "a used module's own use t27c does not splice".

Expected

The refusal is decided by the module (the last segment of the #[path] string), so any alias is refused.

Found by

Writing the direct tests of path (lane C1 of #10522, issue #10825). The test that shows it is known_defect_the_splice_refusal_reads_the_alias_not_the_module_path in rust_convert.t27: it pins the current behaviour and flips when this is fixed.

User Scenarios

• Given a Rust file with #[path = "../../../gen/rust/compiler/int_type_names.rs"] mod names; and a function calling names::is_integer_type_suffix, when tri t27o converts it, then the conversion must be refused with "a used module's own use t27c does not splice" instead of emitting use compiler::int_type_names; and the spliced call.

Requirements

• FR-001: The splice refusal check in rust_convert.t27 MUST inspect the module path from the K_MODPATH owner slots (st[s+2]..st[s+3]) instead of the alias slots (st[s]..st[s+1]).
• FR-002: The check MUST extract the last path segment of the #[path] string and refuse when that segment equals int_type_names, regardless of the mod alias.
• FR-003: The existing pinned test known_defect_the_splice_refusal_reads_the_alias_not_the_module_path in rust_convert.t27 MUST be updated to assert the corrected behaviour.

Success Criteria

• tri t27c specs/tools/t27o/rust_convert.t27 exits 0
• tri t27c specs/compiler/int_type_names.t27 exits 0
• The test known_defect_the_splice_refusal_reads_the_alias_not_the_module_path in specs/tools/t27o/rust_convert.t27 passes (1 test)
• Converting the repro snippet with tri t27o now refuses with "a used module's own use t27c does not splice" (exit 0 with refusal message)

Boundary

• specs/tools/t27o/rust_convert.t27
• specs/compiler/int_type_names.t27