Conversation
a9d4fec to
b2792e1
Compare
b2792e1 to
977ecd0
Compare
|
Many of the ci failures existed previously: https://travis-ci.org/passcod/notify/builds/579460084 |
|
All of the failures are preexisting, the windows stable failure that appears new I'm guessing is a flaky test |
…ther platforms too
|
I'll have to check this when I've got some time. Please ping me again otherwise in 3 days. |
|
I think we want to warn users about the broken state of the debouncer in these cases, if we're not able to fix it sufficiently. This affects at least 4.X stable and 5.X. So breaking the debouncer isn't possible. Your PR also mentions macOS specifically. This issue persists at least also on linux, it may also be another issue with similar symptoms. |
|
The |
- panic on deadlock in delay_zero test. - don't print spammy messages to terminal
737b5d9 to
0b632d7
Compare
|
I pushed up the Testing both platforms again with the deadlock fix shows the |
|
Thanks for all the work. I re-scheduled CI and will verify on linux again. |
|
I can confirm that this resolves the deadlock problem under linux as well. |
|
|
Works around the issue described in #205. Tested on the
fsevent backend onlylinux/macos/windows.I tried to change locking as little as possible as it's somewhat hard to follow, and instead use a simple backoff algorithm to workaround the deadlock.
These tests fail on both branch
origin/try-v4and my fork.Would be nice to get a new version out with the fix as well!