Observed behavior
While implementing integration tests of flagd, I initially ran the tests against the flagd binary which seemingly arbitrarily passed/failed. Conversely, the tests always pass while using the built flagd image directly.
The "flakey" tests are the ones that make a change to the flag config file, sleeps 50 milliseconds & then asserts that the flag has been cleared from cache.
Increasing the sleep (all the way up to 2 seconds) did nothing to fix the flakiness. This leads me to believe the behaviour of my mac's file change notifications system is inconsistent with the behaviour of flagd's distroless image's system.
Expected Behavior
The tests to pass as consistently when using the flagd binary as when using flagd's image.
Steps to reproduce
Not easily reproducible right now (the flagd integration tests and its dependencies are yet to be merged). Will update when they have been
Observed behavior
While implementing integration tests of flagd, I initially ran the tests against the flagd binary which seemingly arbitrarily passed/failed. Conversely, the tests always pass while using the built flagd image directly.
The "flakey" tests are the ones that make a change to the flag config file, sleeps 50 milliseconds & then asserts that the flag has been cleared from cache.
Increasing the sleep (all the way up to 2 seconds) did nothing to fix the flakiness. This leads me to believe the behaviour of my mac's file change notifications system is inconsistent with the behaviour of flagd's distroless image's system.
Expected Behavior
The tests to pass as consistently when using the flagd binary as when using flagd's image.
Steps to reproduce
Not easily reproducible right now (the flagd integration tests and its dependencies are yet to be merged). Will update when they have been