Reenable sleep ns spec on MacOS #1212
Conversation
|
This failure seems to be much more than just unreliability in I'm attempting to disable parallel specs on MacOS to see if that helps this produce more reasonable results. |
|
FYI: #1213 has a change that improves the " Expected false == true" error message to show the actual values. Feel free to integrate that in your branch if that makes debugging easier.. Otherwise, I'll just leave that PR on Draft until this issue is fixed, to prevent adding another merge conflict. |
|
Let's close, it doesn't even pass reliably on Linux |
|
Please don't close my PRs. |
This spec was originally added to confirm that sub-millisecond sleeps actually did sleep; on JRuby and TruffleRuby before fixes a sleep of 0.0001s would immediately return, because we both used a sleep function with a minimum resolution of 0.001s. Ideally 100 sleeps of 0.0001s should not exceed 0.03s, but since that's not the goal of this spec and since it makes the spec flaky under load or on slower systems, it seems best to remove this check. The remaining check just confirms that 100x sleep of 0.0001s does actually sleep for at least 0.01s. Any Ruby failing the spec now would indicate they are not actually sleeping for 0.0001s and they need a fix.
headius
force-pushed
the
sleep_ns_spec_on_macos
branch
from
f1296d9
to
252996c
Compare
|
@eregon I have made the change I suggested in #1207 (comment), removing the |
Investigating why this spec fails on MacOS in hopes of re-enabling it.