solidjs / solid / 41023423747

78%

main: 89%

push

github

test(web): two hydrate specs wait for what they assert, not a fixed number of ticks (#4436)

* test(web): C3 (b) waits for the trace tier's load, not a count of ticks

The arm holds an occurrence on the lazy traces tier and asserted the late
claim after two quiesce() calls. The tier's load is a real import() through
the module runner, which took 7-13 ms locally while the two quiesce() calls
took 7.5-11 ms, so the arm failed whenever the import came in second
(5 of 6 runs on next, macOS). It now awaits the load itself, as
gateTraceTier().release() does, before letting the claim settle. The hold
is still pinned: no fill and hydration in progress right after hydrate(),
and the fill already counted at the hydration-end callback.

* test(web): the late-boundary truncation spec waits for done, not 80 ms

hydrate() plus its flush took 12-218 ms locally on a busy machine, and
done lands two macrotasks after it. When the first of those ran past the
spec's fixed sleep(80), the timer fired first and done was still
undefined (2 of 3 isolated runs failed on next while another job loaded
the machine). Stalling the event loop for 100 ms right after scheduling
the wait reproduces it every time. The spec now polls done every 10 ms
for up to a second; a boundary that waits for a swap instead of the
rejected path still never reaches done and fails.

790 of 1088 branches covered (72.61%)

Branch coverage included in aggregate %.

1125 of 1359 relevant lines covered (82.78%)

29.88 hits per line