[aaltat]

Summary

Tests only, no production code. They pin what Python sends to Node and how it reads the reply for a generated JavaScript extension keyword, before a follow-up changes the generated keyword to call call_js_keyword and lets call_js_keyword fill defaultTimeout when the caller leaves it out (follow-up to #5318, #5321).

generated keyword (Browser._jskeyword_call)
  signature from the JS function   + defaults: NaN, Infinity, -Infinity, false, null, undefined, "str", 3
  arguments sent                   + *args as one "args" list
                                   + defaultTimeout from a Test scoped timeout
  reply read                       + bodyPart chunks joined (split inside a multi-byte character)
                                   + no result -> None

The tests sit at the Python seam: utest/test_js_extension_call.py with a stub that records CallExtensionKeyword. The repeated setup became pytest fixtures.

Not tested on purpose: a reply carrying json. Node builds every extension response with jsonResponse('', ...), so the last_json branch can't be reached; the follow-up removes it.

Evidence

• Before: the stub returned [], so no unit test read a reply; *args, JS defaults and a Test scoped timeout were covered only by acceptance tests or not at all (NaN/Infinity).
  After: pytest utest/test_js_extension_call.py: 9 passed. Breaking the join, the None return, *args, the NaN conversion or the timeout in _jskeyword_call fails the six new tests.

Merge Danger

Door: two-way

Blast Radius: none

Unit tests only; nothing ships to users.

🤖 Generated with Claude Code