types ratchet: 26 new conflicted type names on master, hidden while Corpus ratchet died at its first step #7437

Since 07:23Z on 2026-10-07, the Corpus ratchet job on master has stopped at its first step, "No new assertionless spec test". The cause was assert true tests in two ports (#7211, #6875), so every later step was skipped, including "A type name may not gain a second definition". PR #7397 fixes that first step. Once it lands, the types step runs again and fails.

At 2b74c2b (the #7397 head, clean tree on the t27c lab), tri types ratchet reports ledger 84 name(s), observed 110 and lists 26 NEW conflicts. Every one of them comes from a specs/port/ file merged in the port batch:

• AppState: trios-server mcp_endpoints/experience.t27 vs tasks.t27
• AsmSection, AssembledInstr: specs/fpga/assembler.t27 vs port compiler/AsmRelocEntry_new.t27 and AssembledInstr_r_type.t27
• BlinkState (4) and Inputs (4): trinity/fpga/openxc7-synth blink_*, d6_blink, d_oscillator and d_slow_blink
• Block: scripts/gen_w377.t27 vs tools/dedupe_identical_tests.t27
• Cli: trios-doctor main.t27 vs trios-igla-race main.t27
• FileChange: specs/file/schema.t27 vs trios-git status.t27
• HirBoardResources: compiler/HirIpCatalog_new.t27 vs HirTernaryRegFile_new.t27 (DRIFT: same fields, []const u8 vs str)
• HwType: specs/fpga/hw_types.t27 vs compiler/HwType_hw_width.t27
• IdentityConfig: browseros lib/identity.t27 vs specs/sacred/sacred_identity.t27
• LinkSection, LinkedSymbol: specs/fpga/linker.t27 vs compiler/LinkSection_text.t27
• LockFreeStack: trinity gen_lockfree_stack.t27 vs specs/tri/collections/lockfree_stack.t27
• LockGuard: trios-cli cmd/report.t27 vs lock.t27
• Path: trios-server tools/git.t27 vs specs/tri/io/fs.t27
• Point: trinity gen_transform.t27 vs specs/tri/math/bezier.t27
• RTree: trinity gen_rtree.t27 vs specs/tri/trees/rtree.t27
• RunResult: trios-cli cmd/run.t27 vs trios-trainer train_loop.t27
• Segment, Segments: scripts/gen_w390.t27 vs gen_w391.t27
• TestResults: fpga/verilog gft_smul_jtag, gft_train1_jtag and vivado gf16_uart_sim_bench
• ToolDefinition, ToolRegistry: browseros tools/tool-registry.t27 vs specs/tools/schema.t27
• Value: trios-server mcp_endpoints/tasks.t27, tools/git.t27 and tools/mod.t27
• Variant: trios-macos clade-improve oversight.t27 vs specs/tri/collections/variant.t27

tri types classified also reports 26 UNJUDGED names.

How to fix: rename the port-side struct to a module-unique name when the two types are DISTINCT, which is the #7315 approach (ExitPool, KidTree). Unify them when they DRIFT. Blessing all 26 would make the ratchet meaningless. Per-file full listing: tri types dup.

Related: #7294 (Tree), #7393.

Boundary

• compiler/AsmRelocEntry_new.t27
• compiler/HirIpCatalog_new.t27
• compiler/HwType_hw_width.t27
• compiler/LinkSection_text.t27
• scripts/gen_w377.t27
• scripts/gen_w390.t27
• specs/file/schema.t27
• specs/fpga/assembler.t27
• specs/fpga/hw_types.t27
• specs/fpga/linker.t27
• specs/sacred/sacred_identity.t27
• specs/tools/schema.t27
• specs/tri/collections/lockfree_stack.t27
• specs/tri/collections/variant.t27
• specs/tri/io/fs.t27
• specs/tri/math/bezier.t27
• specs/tri/trees/rtree.t27
• tools/dedupe_identical_tests.t27
• tools/git.t27
• tools/mod.t27
• tools/tool-registry.t27

User Scenarios

• Given the Corpus ratchet job is unblocked by PR specs(port): nine batch-merged ports spell string arrays as [N]str, so gen-verilog lowers them (Closes #7393) #7397, When tri types ratchet runs on master, Then it must report zero new conflicted type names (ledger count equals observed count).
• Given 26 type names currently have duplicate definitions across port and non-port specs, When the port-side structs are renamed to module-unique names or unified where they drift, Then tri types classified must report zero UNJUDGED names.

Requirements

FR-001: Each of the 26 conflicted type names MUST be resolved by either renaming the port-side definition to a module-unique name or unifying the definitions when they represent the same type with drift.
FR-002: The resolution MUST not change the public API of any non-port spec file.
FR-003: After resolution, tri types ratchet MUST exit 0 with ledger count equal to observed count.
FR-004: After resolution, tri types classified MUST report zero UNJUDGED names.

Success Criteria

• tri types ratchet exits 0 and reports ledger 110 name(s), observed 110
• tri types classified reports UNJUDGED: 0
• tri types dup produces no output