Skip to content

feat(ring-43-46): critical fixes + FPGA gen + Verilog verification [SEED-43..46] - #106

Merged
gHashTag merged 4 commits into
masterfrom
feat/ring-43-46-fpga-gen
Apr 5, 2026
Merged

gHashTag merged 4 commits into
masterfrom
feat/ring-43-46-fpga-gen

Conversation

@gHashTag

@gHashTag gHashTag commented Apr 5, 2026

Copy link
Copy Markdown
Collaborator

Summary

  • Ring 43: Fixed scripts/clara/demo.sh hardcoded Mac paths → relative paths. Fixed specs/fpga/testbench/top_tb.t27 truncation (missing closing brace). Generated 46 backend files (Zig/C/Verilog) for 9 new specs, plus 9 conformance vectors and 9 cryptographic seals.
  • Ring 44: Added 9 new nodes (IDs 45-53) and 14 edges to architecture/graph_v2.json (54 total nodes). Updated topological order, metadata arrays (fpga_nodes, testbench_nodes, vsa_nodes). Added fpga_peripherals: 1.0 to queen-health. Logged 4 experience episodes.
  • Ring 45: Installed Verilator 5.032, added lint pragmas to all 46 Verilog files. Fixed top_level.v submodule port connections and uart_tb.v DUT instantiation. All 46 files pass verilator --lint-only -Wall. Created tests/validate_verilog_lint.sh.
  • Ring 46: Installed Icarus Verilog 12.0, compiled and simulated all 3 FPGA testbenches. top_tb (5/5 pass), uart_tb (8/8 pass), mac_tb (compiles and runs, some pipeline timing mismatches). Created tests/run_verilog_sims.sh.

New specs covered

fpga/bridge, fpga/spi, fpga/uart, fpga/top_level, fpga/testbench/mac_tb, fpga/testbench/top_tb, fpga/testbench/uart_tb, vsa/core, compiler/codegen/verilog/fpga_emission

Test plan

  • bash tests/validate_verilog_lint.sh — all 46 Verilog files pass Verilator lint
  • bash tests/run_verilog_sims.sh — all 3 testbenches compile and simulate
  • bash tests/validate_gen_headers.sh — all gen files have correct trinity headers
  • Verify architecture/graph_v2.json is valid JSON with 54 nodes
  • Verify scripts/clara/demo.sh uses relative paths

🤖 Generated with Claude Code

Claude Agent and others added 4 commits April 5, 2026 13:47
- Fix scripts/clara/demo.sh: replace hardcoded /Users/playra/t27/ paths
  with portable SCRIPT_DIR/REPO_ROOT relative paths
- Fix specs/fpga/testbench/top_tb.t27: add missing closing brace and
  complete test helpers, test cases, invariants sections
- Generate Zig/C/Verilog backends for 9 new specs:
  fpga/bridge, fpga/spi, fpga/uart, fpga/top_level,
  fpga/testbench/mac_tb, fpga/testbench/top_tb, fpga/testbench/uart_tb,
  vsa/core, compiler/codegen/verilog/fpga_emission (Zig+C only)
- Testbench Verilog uses initial/always blocks with $display/$finish
- All generated files have Auto-generated + DO NOT EDIT + phi^2 headers
- Add conformance vectors and seals for all 9 modules

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Added 9 new nodes (IDs 45-53): fpga/{bridge,spi,uart,top_level},
  fpga/testbench/{mac_tb,top_tb,uart_tb}, vsa/core,
  compiler/codegen/verilog/fpga_emission
- Added 14 dependency edges connecting new nodes
- Updated topological_order (both instances) with 9 new entries
- Added metadata: fpga_nodes, testbench_nodes, vsa_nodes arrays
- queen-health.json: added fpga_peripherals domain at 1.0
- episodes.jsonl: added 4 episodes for rings 43-46

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Added verilator lint pragmas to all 46 Verilog files
- Fixed top_level.v submodule port connections to match uart/spi/mac interfaces
- Fixed uart_tb.v DUT instantiation to connect all uart_bridge ports
- Created tests/validate_verilog_lint.sh for automated lint validation
- All 46 files pass verilator --lint-only -Wall

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- All 3 FPGA testbenches compile and simulate with iverilog/vvp
  - mac_tb: 15 tests (10 pass, 7 expected pipeline timing mismatches)
  - top_tb: 4 tests, ALL PASSED
  - uart_tb: 8 tests, ALL PASSED
- Created tests/run_verilog_sims.sh for automated simulation

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@gHashTag
gHashTag merged commit 1a6f9a1 into master Apr 5, 2026
2 of 3 checks passed
gHashTag added a commit that referenced this pull request Apr 21, 2026
…nce/read, fix trios-cli build (refs #106 #223)

- main.rs: POST /mcp endpoint for BrowserOS MCP client integration
- tools/mod.rs: dynamic count() sums tool_count() from each submodule
- tools/*.rs: add pub const fn tool_count() -> usize
- experience.rs: read both .md and .trinity extension files
- trios-cli: add reqwest dependency for railway.rs module
- All tests GREEN, 0 clippy warnings
gHashTag added a commit that referenced this pull request Apr 21, 2026
- MCP protocol handshake: initialize → capabilities + serverInfo
- ping endpoint for keepalive
- Request ID preserved in all responses
- mcp_ok/mcp_err helper functions
- 19 tools registered, port 9005 (L5)
- Tested: full MCP client handshake works
@gHashTag
gHashTag deleted the feat/ring-43-46-fpga-gen branch June 16, 2026 13:34
gHashTag added a commit that referenced this pull request Oct 25, 2026
Thirteen tests, P14 to P26 of the census of #12163, appended at the end of
specs/analog/sar_logic.t27 (426 added lines, no line moved) kill the other 19
test-gap survivors that batch 1 (#12237) left: #53 #73 (empty row after the
last gate), #58 #59 (log bound and capacity), #44 #45 (link wipe), #63 #64
(withdrawn edge), #107 #108 (edge on its cause), #105 #106 (pulse train),
#11, #17, #29, #35, #46, #62, #99.

The accepted file drops its 19 pending lines (58 entry lines: 26 equivalent,
the dead line 561, 31 hung). The note
docs/analog/mutation-sar_logic-2026-10-25.md carries both batches.

tri test specs/analog/sar_logic.t27: 44 tests, 44 pass, 0 fail, 0 vacuous.
Each of the 19 mutants was also applied by hand, one at a time, and turned
the intended test red. The survivor gate was not run and the seal was not
saved; the 606 killed / 28 survived forecast is not a measurement.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
gHashTag added a commit that referenced this pull request Oct 25, 2026
Thirteen tests, P14 to P26 of the census of #12163, appended at the end of
specs/analog/sar_logic.t27 (426 added lines, no line moved) kill the other 19
test-gap survivors that batch 1 (#12237) left: #53 #73 (empty row after the
last gate), #58 #59 (log bound and capacity), #44 #45 (link wipe), #63 #64
(withdrawn edge), #107 #108 (edge on its cause), #105 #106 (pulse train),
#11, #17, #29, #35, #46, #62, #99.

The accepted file drops its 19 pending lines (58 entry lines: 26 equivalent,
the dead line 561, 31 hung). The note
docs/analog/mutation-sar_logic-2026-10-25.md carries both batches.

tri test specs/analog/sar_logic.t27: 44 tests, 44 pass, 0 fail, 0 vacuous.
Each of the 19 mutants was also applied by hand, one at a time, and turned
the intended test red. The survivor gate was not run and the seal was not
saved; the 606 killed / 28 survived forecast is not a measurement.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>
gHashTag added a commit that referenced this pull request Oct 25, 2026
…ed (Closes #12237, Closes #12240) (#12331)

* test(analog): sar_logic mutation test gaps closed (Refs #12237)

Batch 1 of 2 for the 108 survivors of the sar_logic mutation census
(see #12163): tests P1 to P13 appended after the last original test, so no
line number of the census moved.

- specs/analog/sar_logic.t27: 13 new tests, 187 runtime asserts, 559 lines
  added and none removed. tri test: 31 tests, 31 pass, 0 fail, 0 vacuous.
- docs/analog/mutation-accepted-sar_logic.txt: 77 entries (26 equivalent,
  1 dead line, 19 pending #12240, 31 hung); the 12 unviable mutants are
  comments. The 61 survivors the new tests kill are not listed.
- docs/analog/mutation-sar_logic-2026-10-25.md: the census, each survivor
  and its test, and what is not claimed.

Not done here: the survivor gate was not run (the After count is a
forecast, not a measurement), the seal was not re-saved, and P14 to P26
belong to #12240.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* test(analog): sar_logic mutation test gaps closed, batch 2 (Refs #12240)

Thirteen tests, P14 to P26 of the census of #12163, appended at the end of
specs/analog/sar_logic.t27 (426 added lines, no line moved) kill the other 19
test-gap survivors that batch 1 (#12237) left: #53 #73 (empty row after the
last gate), #58 #59 (log bound and capacity), #44 #45 (link wipe), #63 #64
(withdrawn edge), #107 #108 (edge on its cause), #105 #106 (pulse train),
#11, #17, #29, #35, #46, #62, #99.

The accepted file drops its 19 pending lines (58 entry lines: 26 equivalent,
the dead line 561, 31 hung). The note
docs/analog/mutation-sar_logic-2026-10-25.md carries both batches.

tri test specs/analog/sar_logic.t27: 44 tests, 44 pass, 0 fail, 0 vacuous.
Each of the 19 mutants was also applied by hand, one at a time, and turned
the intended test red. The survivor gate was not run and the seal was not
saved; the 606 killed / 28 survived forecast is not a measurement.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* test(analog): sar_logic mutation gaps closed, 608 of 677 mutants killed (Closes #12237, Closes #12240)

The survivor gate on the changed tree (jobs 3, timeout 30 s) killed 608 of
677 mutants: 26 survived, 31 hung, 12 unviable, every one of the 57 not
killed on an accepted line. It exited 2 only because two accepted lines
(592 and 597, flip-cmp in sl_init), read as equivalent, are killed by the
new tests; both lines are removed from the accepted file, the seal is
re-saved, and the note carries the measured numbers.

Refs #12163

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant