Migrate bus primitives to `amaranth.lib.wiring`

[jfng]

Supersedes #48, as the changes to CSR and Wishbone are similar in nature, and would benefit from a common review.

[jfng]

Should we make path= a part of our vocabulary? path is kind of a Catherine-ism, just what I used for years to refer to this sort of tuple. It's entirely possible another name is better.

In the latest changes, I have kept path= in CSR and Wishbone Interface.__init__(), as they are wrappers of lib.wiring.Interface with additional metadata.

However, I have removed path= from Decoder and Arbiter.__init__(). The name of their bus interfaces is an implementation detail. This is consistent with how things were before this PR.

[whitequark]

@jfng

In the latest changes, I have kept path= in CSR and Wishbone Interface.__init__(), as they are wrappers of lib.wiring.Interface with additional metadata.

Thanks. My question is actually broader in scope than just SoC; I wonder if "path" is really a good term for this in general, including Amaranth itself.

[jfng]

Thanks. My question is actually broader in scope than just SoC; I wonder if "path" is really a good term for this in general, including Amaranth itself.

I prefer "path" over alternatives like "name", as it conveys the idea that it is a sequence of strings, rather that a string itself.

Amaranth SoC uses "name" for a similar situation (resource names are prefixed by names of parent windows when iterating MemoryMap.all_resources()), and I think "path" would be appropriate here too.

[jfng]

The following test fails with Python 3.8:

FAIL: test_trigger_wrong (tests.test_event.SourceSignatureTestCase)
----------------------------------------------------------------------
ValueError: 'foo' is not a valid Trigger

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/runner/work/amaranth-soc/amaranth-soc/tests/test_event.py", line 54, in test_trigger_wrong
    src = event.Source.Signature(trigger="foo")
AssertionError: "'foo' is not a valid Source.Trigger" does not match "'foo' is not a valid Trigger"

But passes with Python 3.11.5

[jfng]

The following test fails with Python 3.8:

FAIL: test_trigger_wrong (tests.test_event.SourceSignatureTestCase)
----------------------------------------------------------------------
ValueError: 'foo' is not a valid Trigger

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/runner/work/amaranth-soc/amaranth-soc/tests/test_event.py", line 54, in test_trigger_wrong
    src = event.Source.Signature(trigger="foo")
AssertionError: "'foo' is not a valid Source.Trigger" does not match "'foo' is not a valid Trigger"

But passes with Python 3.11.5

This error message format was changed in Python 3.9.

For consistency across supported Python versions, we can raise our own ValueError instead.

[jfng]

In the latest changes, Interface constructors take the same parameters as their corresponding Signatures (in addition to path=)`.

[jfng]

Merged. Thanks @whitequark for the review!