meshbay.git/packages/meshbay-hub/tests/test_offer_retry.py, branch 0.17

test: make both suites pass on Windows

2026-09-27T20:21:26Z

Most of these failed on Windows for reasons that had nothing to do with the code under test, which is how real Windows defects hid among them: - Read and write files as UTF-8, and talk to Node in UTF-8. read_text(), write_text() and subprocess text=True use the locale codepage, cp1252 on Windows: "é", "—" and "→" arrived as "?" or crashed, some sixty tests. Calls to PowerShell and schtasks are left alone -- they answer in the console codepage. - Import ESM harness modules by file URL (as_uri): a raw "C:\..." path is not a module specifier. - test_cli_golden: mask the tmp path in its JSON-escaped form, spell it the POSIX way, record on Linux, mask the protocol version (the recording had failed everywhere since the MNP 4.0 bump) and argparse's version-dependent quoting; point USERPROFILE at the tmp home, or `member invite` and `operator pair` wrote their codes into the developer's profile. - test_disk_io_off_loop: expect what a free loop can reach on the platform's timer, 15.6 ms on Windows, not an assumed 5 ms. - test_root_paths_are_operator_only: expect the OS's spelling of the path. - test_audio_meta_cache: find ffprobe with shutil.which. Node suite on Windows: 1489 passed, none failed. Hub suite: 3 failures left, all older than this change (two SQLite concurrency tests, one transfer resume). Co-Authored-By: Claude Opus 5.5

test(hub): read the transport wherever it is split

2026-09-25T10:07:49Z

spa_source.transport_files() takes the classic transport*.js scripts from the hub's shell, in load order. Every test that read transport.js reads them all, the Node harnesses run them joined as one scope, the chat probe loads each, and the desktop shell must load them in order. Co-Authored-By: Claude Opus 5.5

fix(hub): Search reaches each group with one offer, and a busy hub is not a dead node

2026-09-23T22:10:41Z

The other half of the 4G failure. The page negotiated every group twice: the sweep opened a connection, read the index and closed it, then the warm-up opened the same group again. The sweep, the warm-up and the tiles each had a concurrency ceiling of their own, and together they went past what the hub admits per account. Whatever the hub refused was then reported as "node unreachable" and remembered as down, which put that group last next time. - Every connection goes through ConnectionPool, which holds the page's one ceiling (six at once, sized for twenty groups on a phone) and keeps what the sweep opened for the tiles. A visit costs one offer per group. A refresh costs none for a connection that answers a four-second ping, and a connection that died while the phone slept is replaced, not waited on. - A connection whose index is being read is held against eviction. With more groups than the pool keeps, it was otherwise the least recently used one. - Negotiations still under way when the page closes close what they get, and a sweep cut short that way remembers nobody as down. - transport.js sends an offer again on 429, 502 or 503, honouring Retry-After, with jittered waits of about twenty seconds at worst. Search counts each retry as progress. A 404, 403 or 504 still fails at once, so a dead node costs no time. The fan-out tests assumed a ceiling of three and were re-measured: four dead groups of twelve now hold nothing back, even on a first visit. The pool and the retry run as shipped code, lifted as text, against a fake clock. Each guard was checked by removing it and seeing its test fail. Co-Authored-By: Claude Opus 5.5