[cmj0121]

Closes #359.

What

assert xs == ys over two lists builds, and a failing one reports both sides:

AssertionError: a.zg:11  assert words(1) == words(2)
  words(1) = ["1", "a, b"]
  words(2) = ["2", "a, b"]

The claim was never the problem. An assert's message is an f-string with a hole per operand, a hole lowers
through str(…), and str(…) refused any list that is not bytes or code points (E4011, "a composite has no
rendering of its own"). Every composite has rendered since the structural rendering was built — print xs and
xs.display() used it — and that arm was the one place still turning a list away. It now keeps its two
bridges and hands every other list to the same renderer.

So this is wider than the issue's title: str(xs) and f"{xs}" render a list as [1, 2], which is what
docs/runtime/format.md already said all four spellings do. E4011 retires.

Decisions made here

• Fixed at the cause rather than at assert. The issue offered a narrower way out (keep the refusal, reword
  it for assert). The chapter says a list renders through a hole, so the refusal was the deviation.
• The two bridges are unchanged. str(bs) and f"{bs}" over a list[byte] or list[rune] are still the
  text, while print bs shows [104, 105]. That difference predates this change; format.md now says so.

Found on the way, not fixed

• A channel INSIDE a composite renders as its address: print [ch], f"{m}" for a map holding one, a tuple.
  Top level is E4076. This is on main today through print; this change adds the list-in-a-hole spelling to
  the ways of reaching it. Not filed yet.
• c_renders_as_text still answers "no" for a list that does not bridge, as it does for a map, so the
  xs.format("04d") refusal says there is no rendering to fall back on when there is one.

Verified

• Local gates, one at a time, each read: build fixpoint lint fmt-self error-codes-check seed-gaps refuse reject
  mem-check oracle corpus sanitize-corpus mem-peak examples docs-mirror docs-links docs-zerg docs-refused
  docs-repeat marker-codes chapter-codes deviation-check refusal-cites method-gaps gates treesitter dead-code
  stdlib-test compiler-test check-equal reject-fuzz fmt-corpus behaviour conformance productions suites desugar.
  refuse and mem-check also under /bin/bash.
• The new mem-check case was red with the renderer's give-back skipped for a list (2730 per round), green
  after.
• The corpus case leaks the message of an Err a guard caught, the class assert_claim is already listed
  for; it is listed beside it under the same ?. With the guarded asserts removed it reports no leak.

Not verified

• make test as one board was not run. make lsp was not run (zrt_listdir sorts by insertion: a large directory costs its square, and make lsp fails under a crowded $TMPDIR #361).

🤖 Generated with Claude Code

https://claude.ai/code/session_01N8jd1oK2orM6diHxhvWhC7