Skip to content

Nine specs whose tests do not compile (blocked) cannot be sealed #7641

Open

Description

@dmitrii-f-t27

Goal

Nine specs added by merged PRs do not get a seal that proves anything, because their tests do not compile with Zig 0.16.0 (t27c seal records blocked). Each needs its generated test code to compile so that the seal can record passed == total > 0.

spec first compile error
specs/compiler/rust_repeat_counts.t27 incompatible types: 'struct { u32, u32, u32, u32 }' and 'comptime_int'
specs/port/fpga/verilog/mvp_ternary_classifier_jtag_noport.t27 invalid digit 'u' for hex base
specs/port/fpga/vivado/gf16_matmul_top.t27 incompatible types: 'bool' and 'u32'
specs/port/tools/check_conflict_markers.t27 expected type expression, found '&'
specs/port/tools/jtag/link_relay.t27 expected type 'u8', found 'u32'
specs/port/tools/jtag/mpsse_jtag.t27 unused function parameter
specs/port/tools/jtag/read_user1.t27 value of type '@typeof(undefined)' ignored
specs/port/tools/jtag/read_verdict.t27 use of undeclared identifier 'Mpsse'
specs/ui/viewport.t27 overflow of integer type 'u8' with value '648'

Success Criteria

Each spec either compiles under t27c test-report and is sealed with passed == total > 0, or the issue names the compiler defect it is blocked on. Why it matters: the Queen map counts a closed issue as covered only when its specs carry such a seal.

Scope note

Spec or generator fixes only; no gate changes.

User Scenarios

  • Given nine specs whose generated test code fails to compile under Zig 0.16.0, When t27c seal is run on each spec, Then the seal records blocked instead of passed == total > 0

Requirements

  • FR-001: Each of the nine cited specs MUST have its generated test code compile under Zig 0.16.0 via t27c test-report
  • FR-002: Each spec MUST achieve passed == total > 0 when sealed via t27c seal
  • FR-003: Fixes MUST be limited to spec or generator changes; no gate changes

Success Criteria

  • t27c test-report specs/compiler/rust_repeat_counts.t27 compiles and t27c seal specs/compiler/rust_repeat_counts.t27 exits 0 with passed == total > 0
  • t27c test-report specs/port/fpga/verilog/mvp_ternary_classifier_jtag_noport.t27 compiles and t27c seal specs/port/fpga/verilog/mvp_ternary_classifier_jtag_noport.t27 exits 0 with passed == total > 0
  • t27c test-report specs/port/fpga/vivado/gf16_matmul_top.t27 compiles and t27c seal specs/port/fpga/vivado/gf16_matmul_top.t27 exits 0 with passed == total > 0
  • t27c test-report specs/port/tools/check_conflict_markers.t27 compiles and t27c seal specs/port/tools/check_conflict_markers.t27 exits 0 with passed == total > 0
  • t27c test-report specs/port/tools/jtag/link_relay.t27 compiles and t27c seal specs/port/tools/jtag/link_relay.t27 exits 0 with passed == total > 0
  • t27c test-report specs/port/tools/jtag/mpsse_jtag.t27 compiles and t27c seal specs/port/tools/jtag/mpsse_jtag.t27 exits 0 with passed == total > 0
  • t27c test-report specs/port/tools/jtag/read_user1.t27 compiles and t27c seal specs/port/tools/jtag/read_user1.t27 exits 0 with passed == total > 0
  • t27c test-report specs/port/tools/jtag/read_verdict.t27 compiles and t27c seal specs/port/tools/jtag/read_verdict.t27 exits 0 with passed == total > 0
  • t27c test-report specs/ui/viewport.t27 compiles and t27c seal specs/ui/viewport.t27 exits 0 with passed == total > 0

Boundary

  • specs/compiler/rust_repeat_counts.t27
  • specs/port/fpga/verilog/mvp_ternary_classifier_jtag_noport.t27
  • specs/port/fpga/vivado/gf16_matmul_top.t27
  • specs/port/tools/check_conflict_markers.t27
  • specs/port/tools/jtag/link_relay.t27
  • specs/port/tools/jtag/mpsse_jtag.t27
  • specs/port/tools/jtag/read_user1.t27
  • specs/port/tools/jtag/read_verdict.t27
  • specs/ui/viewport.t27

Activity

  1. dmitrii-f-t27 commented on Oct 10, 2026

    @dmitrii-f-t27
    ContributorAuthor

    Taking the nine-spec test/seal audit on branch codex/spec-tests-7641, based on master 606611ced01f3970553dced70e1a2547ab8fb383. I will reproduce each failure with a compiler built from this tree, preserve each spec's behavior, and record actual generator blockers instead of removing meaningful tests. Scope: the nine paths in this issue and their generated seals; no gate changes.

    Dmitrii Fedorov (@dmitrii-f-t27)

  2. added a commit that references this issue on Oct 11, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions