[Build Fix] Updates for modern Cocotb API and Verilator Linting (macOS)

[muditbhargava66]

Hi! I recently pulled the repo to build and simulate the design on macOS. I encountered a few build failures due to tool version differences (specifically deprecated Cocotb syntax and stricter Verilator linting).

I have applied the following fixes to get the environment running. I wanted to document them here in case you'd like a PR, or for future users facing similar build issues.

1. Linting Fixes (Verilator)

• File: src/io_inft.v
  • Issue: Unused signal data_v_early.
  • Fix: Added /* verilator lint_off UNUSEDSIGNAL */ to suppress the warning and allow clean linting.

2. Testbench Fixes (Cocotb Compatibility)

It appears I am running a newer version of Cocotb than originally used, which caused deprecation errors.

• File: test/Makefile
  • Issue: Variable COCOTB_TEST_MODULES was deprecated/ignored.
  • Fix: Changed to MODULE = test.
• File: test/test.py
  • Issue 1: Deprecated Clock(..., unit="us").
  • Fix: Updated to Clock(..., units="us").
  • Issue 2: Deprecated BinaryValue.to_unsigned().
  • Fix: Updated to use .integer property.

3. Verification Results:

• C Reference (make tv): Passed.
• RTL Lint (make lint): Passed (1 warning fixed).
• RTL Simulation (make test): Passed 3/4 tests.
  • hash_spec_test : PASS (Matches spec vectors)
  • hash_test : PASS (Randomized testing)
  • slow_output_test : PASS
  • dissable_test : FAIL (Likely due to uninitialized 'X' states during disable check)

Question: Is the dissable_test failure a known sensitivity with specific reset timing, or should I be looking into the reset logic for the X-state propagation?

[Essenceia]

Hi,
Thanks a lot for taking a look at this repo and finding all these issues.
I am currently traveling so it is going to take me a few days to address this.

Could you please specify which version of cocotb are you using, so I can properly reproduce these issues.

As for actually addressing the cocotb version specific syntax, I might have to see if the syntax is backwards compatible and see how flexible the tinytapeout flow is with supporting more up to date versions. If the tinytapeout flow expects a fixed cocotb version, I am afraid I might need to impose there cocotb version.

As for your question, depending on the order the tests where run there could well be a sensitivity with an X prop causing this test to fail. I will look into it. For context, this test was originally designed early on in the design timeline with the intention of checking this block's logic was stable in case it wasn't enabled to help reduce it's dynamic power consumption. Latter, I learned that this block would actually be power gated when not enabled, so this test partially fell into deprecation, thus it is quite possible I let a corner case like this slip. Depending on your use case, this should have no impact on your design's functionality

[muditbhargava66]

Hi, thanks for the reply!

To answer your question: I am currently running cocotb 1.8.1, though I also verified that everything works on cocotb 2.0.0 (which matches your test/requirements.txt).

Regarding Version Compatibility

I noticed that test/requirements.txt pins cocotb==2.0.0, but the current syntax in test.py was deprecated in cocotb 1.5+. Here's a summary of the required changes for modern cocotb versions:

Original Syntax	1.8.1 Compatible	2.0.0 Compatible
COCOTB_TEST_MODULES = test	MODULE = test	COCOTB_TEST_MODULES = test
Clock(..., unit="us")	Clock(..., units="us")	Clock(..., unit="us")
.to_unsigned()	.integer	.to_unsigned()

The tricky part is that cocotb 2.0.0 reversed some of the 1.8.x deprecations (e.g., unit vs units). So the cleanest path forward might be to target cocotb 2.0.0 directly, which aligns with requirements.txt.

Regarding dissable_test

You were spot on about the X-propagation sensitivity. I fixed this by using .binstr for comparison instead of integer equality, which handles 'X' states gracefully:

uo_out = dut.uo_out.value.binstr  # Instead of .value
# ...
assert(dut.uo_out.value.binstr == uo_out)  # String comparison

With the lint fix, cocotb syntax updates, and X-state handling, I now have 4/4 tests passing locally on both cocotb 1.8.1 and 2.0.0.

Happy to open a PR with the complete set of fixes if that helps!