fix(ci): read trinet_node_core.v in the TRI-NET baud and node-v2 workflows

[gHashTag]

Fixes #888. Part of #851 (W15).

Change

• ax7203-trinet-baud.yml and ax7203-trinet-node-v2.yml now read fpga/portable/trinet_node_core.v in every read_verilog/iverilog list, and name it in paths:. 5b09159 moved the node logic there on 2026-08-02. ax7203-trinet-fleet.yml and trinet-portability.yml already read it.
• set -o pipefail on the six steps that pipe into tee. A Yosys error now fails the Yosys step instead of surfacing 16 nextpnr attempts later as no clean seed across heap+sa x8. The Yosys and fasm2bit steps carry a # tee-pipefail: marker, because the gate takes -nocarry and docker … bash -c for a shell of their own.
• tools/tee_pipefail_baseline.txt: the six entries are removed, 78 → 72. The gate is a ratchet, and a stale entry fails it.

Measured

Locally, sources from origin/main 57de4a1:

check	without the core	with the core
iverilog -g2012, Node v2 golden TB	exit 3, Unknown module type: trinet_node_core (the error in CI run 30796536505)	exit 0
vvp	—	SIM RESULT: 6/6 keyed receipts bit-exact (fails=0), TB PASS
yosys hierarchy -check -top trinet_node_v2_ax7203 with the Xilinx cell libraries	exit 1, the error in CI run 37225332379	exit 0
tools/check_tee_pipefail.py	—	122 of 122 workflows read, 72 known, 0 new, exit 0
same, with the Yosys step's marker removed	—	flags the step (the marker is needed)

CI: both workflows have no pull_request trigger, so they were dispatched on this branch at e2756f9. Both passed (2026-10-05):

run	result
Node v2, 37246227964	golden: SipHash TB PASS; SIM RESULT: 6/6 keyed receipts bit-exact (fails=0), TB PASS. bitstream: Yosys 1489 LCs, routed with heap/1, .bit sha256 77b5bfa9…0948e
Baud Ladder, 37246230148	baud30: 1498 LCs, heap/1, .bit bda1ef35…83ce. baud60: 1525 LCs, heap/1, 6d2a8715…a429. baud120: 1545 LCs, heap/1, 6fdf7192…8532

Main's last runs of these two workflows (37225332354, 37225332379) failed at iverilog and after 16 nextpnr attempts respectively.

This PR's doc-refs and orphan-artefacts checks are red for the same files as main since #510 merged (research/arxiv_gft/*, research/GFT_DEFECTS_2026-08-08.md, research/GF_INVENTORY_2026-08-08.md). It inherits them from main and does not touch them. #872 fixes them.

Not measured

• Whether the designs meet 50 MHz. Both workflows run nextpnr with --timing-allow-fail and keep its log in /tmp, so a routed bitstream says nothing about timing.
• All four chipdb cache lookups missed (Cache not found for input keys: chipdb-xc7a200tfbg484-2-regymm-…), and each job built the chipdb. There was no chipdb entry on main to restore: see ci: setup-zig caches fill the 10 GB Actions cache and evict every chipdb #844.
• No board load. None is needed to merge this.

🤖 Generated with Claude Code

[github-actions]

X Brain Health Check

Score: 100.0/100
Status: 🟢 HEALTHY
Threshold: 80/100

X Brain is above merge threshold

[github-actions]

🧪 Brain Stress Test Results

Score: not measured (no stress test prints a Score: line)
Status: 🟢 PASS
Threshold: 270/300

X Stress test passed

[tri-claw]

Summary by t27-bees

This PR fixes two TRI-NET CI workflows to include fpga/portable/trinet_node_core.v (added in 5b09159) in their Verilog source lists and path triggers, and adds set -o pipefail to six steps that pipe into tee so Yosys errors fail fast instead of surfacing later as nextpnr seed failures. The baseline file is updated to remove the now-covered entries.

Worth a look

• .github/workflows/ax7203-trinet-node-v2.yml:30 — The iverilog command lists trinet_node_v2_ax7203.v before trinet_node_core.v and trinet_siphash24.v. If trinet_node_v2_ax7203.v references modules defined in those later files, iverilog may still resolve them (it reads all sources), but the conventional dependency order would be siphash → core → node → testbench.
• .github/workflows/ax7203-trinet-baud.yml:56 — The read_verilog order is trinet_siphash24.v, trinet_node_core.v, trinet_node_v2_ax7203.v; this matches the dependency chain and is correct.
• .github/workflows/ax7203-trinet-node-v2.yml:27 — The SipHash iverilog step adds set -o pipefail but does not include trinet_node_core.v (not needed for that testbench), which is fine.
• .github/workflows/ax7203-trinet-baud.yml:100 and .github/workflows/ax7203-trinet-node-v2.yml:99 — The fasm2bit steps add set -o pipefail with a comment explaining the bash -c shell; the pipe is only around the docker run … | tee, so pipefail will catch a failure from the docker command itself.
• tools/tee_pipefail_baseline.txt — Six entries removed (lines 52–53, 59–62 in the old file). The gate is a ratchet; if any of those steps regress and lose pipefail, the check will fail.

---

_{Read: 3 files, +15 -9, the whole diff. Head 25b0839. Model: openai-compatible/nvidia/nemotron-3-ultra-550b-a55b. Commands: @t27-bees review | summary | help | pause | resume. Free for every repository.}