Conversation
rescript
@rescript/darwin-arm64
@rescript/darwin-x64
@rescript/linux-arm64
@rescript/linux-x64
@rescript/runtime
@rescript/win32-x64
commit: |
|
|
||
| exit_watcher() { | ||
| local watcher_pid="$1" | ||
| rm -f lib/watch.lock |
There was a problem hiding this comment.
Hmm why did this just remove the lockfile, that should be removed when killed
There was a problem hiding this comment.
Ah deleting the lock file would kill the watcher. Is that not the case anymore. Why do we need this complex setup. The pid is also in the lockfile, so we shouldn't need to pass it. But if it behaves correctly it should also exit so we shouldn't need to kill it. Maybe we need to await that it's really closed though? not sure though. At least killing shouldn't be needed, that's exactly what this test is testing. If it fails the behavior is wrong.
There was a problem hiding this comment.
@jfrolich Yeah there may have been some hallucinations here, sorry! Please have another look.
Now just waiting for the pid to disappear after removing the watcher lock file.
Summary
watch.lock.Why
CI was intermittently failing with corrupted compiled interface artifacts while running rewatch tests. The root causes were lock/test races:
watch.lockbut did not wait for the watcher process to terminate before the next test started.