You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
zig-test execution level: 37/69 tri-net gens pass; three finding classes in the rest #1928
Running zig test on tri-net's committed gen/zig (a level above ast-check) with two emission fixes in the linked PR (dead types.zig import skipped when unreferenced; test asserts emit @panic instead of @compileError, which fired whenever a runtime-condition branch was ANALYZED): 37/69 files compile AND pass their test blocks.
The 32 remaining split into three classes, all REAL findings the execution level exists to catch:
The t27 saturation idiom is Zig-fatal: let sum = a + b; if (sum < a) ... panics on overflow at the + (checked semantics) before the guard runs -- specs need +% for the probe add (tri_challenge, timer, packet_queue...). Same class as the tri_sha256 +% migration.
Genuine test failures at runtime (e.g. etx 'dead link should be infinite') -- semantic bugs invisible to every syntax-level gate.
Small emission tails: unused test-binding consts when the assert folds away (hello), a stray @compilelog (wire).
tri-net gates the passing 37 in CI and ratchets as classes 1-3 clear.
User Scenarios
Given the tri-net gen/zig suite at execution level, when zig test runs on all 69 files, then 37 pass and 32 fail across three finding classes (saturation overflow, semantic runtime failures, emission tails)
Given a spec using the saturation idiom let sum = a + b; if (sum < a) ..., when compiled to Zig with checked arithmetic, then the + panics before the guard runs unless the spec uses +% for the probe add
Given specs with emission tails (unused test-binding consts, stray @compileLog), when the assert folds away, then the emitted Zig contains dead code that fails compilation or clutters output
Requirements
FR-001: Specs for tri_challenge, timer, packet_queue, and tri_sha256 MUST replace saturation + with +% so the probe add wraps instead of panicking
FR-002: Specs for etx (and any component with genuine runtime test failures) MUST correct the semantic bug so the test assertion holds at runtime
FR-003: Specs for hello and wire MUST remove unused test-binding consts and stray @compileLog so emission is clean when asserts fold away
FR-004: The t27 compiler MUST continue to gate the passing 37 files in CI and ratchet coverage as classes 1–3 clear
Success Criteria
zig test on tri-net gen/zig executes all 69 files and exits 0
69 test files compile AND pass their test blocks (up from 37)
Zero saturation overflow panics in probe adds across tri_challenge, timer, packet_queue, tri_sha256 specs
Zero runtime assertion failures in etx and related components
Zero unused const or @compileLog emission tails in hello and wire specs
REAL-failures list fully cleared by tri-net#331: all 7 specs (area_optimization, etx, mesh_protocol_stack, network_analytics, performance_benchmarks, self_healing, timing_closure) now pass zig test end-to-end; tri-net ratchet raised 39 -> 46. Remaining scope of this issue: the 5 Zig emission classes.
Reopened 2026-10-07: the bulk close of issues without a ## Boundary section was reverted by the owner. Nothing to do here; this comment replaces the earlier close notice.
Activity
gen-zig: skip dead base-types import; test asserts panic instead of c…
gHashTag commentedon Aug 8, 2026
REAL-failures list fully cleared by tri-net#331: all 7 specs (area_optimization, etx, mesh_protocol_stack, network_analytics, performance_benchmarks, self_healing, timing_closure) now pass zig test end-to-end; tri-net ratchet raised 39 -> 46. Remaining scope of this issue: the 5 Zig emission classes.
gHashTag commentedon Oct 7, 2026
Reopened 2026-10-07: the bulk close of issues without a
## Boundarysection was reverted by the owner. Nothing to do here; this comment replaces the earlier close notice.