[dtt-github]

What

A failing example's account in What actually happened now ends with a Why wasn't it N? link. It lists every line of the design that could have given the signal the expected value, and what happened to it the last time it could have run:

• Blocked: "Line 20 of blinker.v: led <= ~led. Didn't run for cycle 3 (!rst_n was false, !en was false) because cnt == N - 1 was false (cnt = 2)."
• Ran: it ran with the values shown and computed something else, or a later line changed the value.
• Hadn't run yet: nothing had reached its decisions by then.

Lines that only write other constants are left out and counted. If no line can write the expected value, it says so. Values in the answer can be clicked to continue in the existing "Why?" chain, and the answer goes with a request to the AI to fix the design or example. It runs in the browser, with no AI and no cost.

How

• wafer-sim: records every if/case decision with the branch taken (SimResult::decisions, capped at MAX_DECISIONS, with decisions_cut) and maps branches to decisions (cover_decision). Origin::writes names the signal an assignment writes, and Cond::point names its branch.
• wafer-sim-wasm / worker: pass the decision log as typed arrays.
• apps/web/src/lib/why.ts: whyNot() finds the candidate lines, walks their decisions at the last moment they were made, and reports the first one that went another way.

Checked

• Simulator test: the up/down counter records the right branch per clock edge, in time order, plus writes.
• In the browser, blinker with a check made one cycle early: the panel names cnt == N - 1 (cnt = 2), and clicking cnt = 2 continues the "Why?" chain. The test example was restored afterwards.
• UART receiver: the case decision is reported ("state matched one of the listed values (state = 2)").
• Local checks: fmt, clippy -D warnings, workspace tests, web typecheck and build.

Docs: user guide (design view), architecture, CHANGELOG.

Initiated by DigitalTwin Technology GmbH.

🤖 Generated with Claude Code