native-platform
native-platform copied to clipboard
Fix `watch_all_parents: error trying to add kqueue for fd` warnings
When running the test
task on master
(ca15a6581ba2e91ef72510032604a653c783c2f6), the following output occurs:
2020-02-26 08:46 java[94718] (FSEvents.framework) process_dir_events: watch_all_parents: error trying to add kqueue for fd 52 (/private/var/folders/y8/lns4s7jn3jldwky0nm3dmx9m0000gn/T; Bad file descriptor)
2020-02-26 08:46 java[94718] (FSEvents.framework) process_dir_events: watch_all_parents: error trying to add kqueue for fd 47 (/private/var/folders/y8/lns4s7jn3jldwky0nm3dmx9m0000gn; Bad file descriptor)
2020-02-26 08:46 java[94718] (FSEvents.framework) process_dir_events: watch_all_parents: error trying to add kqueue for fd 51 (/private/var/folders/y8; Bad file descriptor)
2020-02-26 08:46 java[94718] (FSEvents.framework) process_dir_events: watch_all_parents: error trying to add kqueue for fd 53 (/private/var/folders; Bad file descriptor)
2020-02-26 08:46 java[94718] (FSEvents.framework) process_dir_events: watch_all_parents: error trying to add kqueue for fd 54 (/private/var; Bad file descriptor)
2020-02-26 08:46 java[94718] (FSEvents.framework) process_dir_events: watch_all_parents: error trying to add kqueue for fd 55 (/private; Bad file descriptor)
2020-02-26 08:46 java[94718] (FSEvents.framework) process_dir_events: watch_all_parents: error trying to add kqueue for fd 54 (/private/var/folders; Bad file descriptor)
2020-02-26 08:46 java[94718] (FSEvents.framework) process_dir_events: watch_all_parents: error trying to add kqueue for fd 47 (/private/var; Bad file descriptor)
2020-02-26 08:46 java[94718] (FSEvents.framework) process_dir_events: watch_all_parents: error trying to add kqueue for fd 52 (/private; Bad file descriptor)
We should understand what is going on there.
See https://e.grdev.net/s/6ma7wj7x5i2bu/console-log?anchor=72.
FTR, I can't reproduce this locally on macOS Catalina, and Google hasn't turned up anything relevant with a quick search.
This seems to happen only for this test: can detect #removedAncestry removed
It doesn't seem to happen if I only run one of the elements of the where
clause - I seem to need to run at least two of them.
Note that the log does not appear as part of the test output, but on the console as build output.
Can you still reproduce this?
With latest master
, the problem seems to be gone.
The issue seem to happen to me again. Re-opening.
We should check if the error happens on our macOS agents. The message seems to end up in the build scan log for the fileWatching:integTest
task.
Maybe this is worth looking at: https://lists.apple.com/archives/filesystem-dev/2010/May/msg00007.html
We don't have proof currently that this happens anywhere else but on @wolfs's machine. I see two ways to move forward:
- try harder to look for the same problem on other machines (like, search for the message in daemon logs on CI etc.),
- try to figure out what's special about Stefan's machine; one idea there is that there's some issue with the file system, so running
fsck
or first aid in Disk Utility could provide some interesting input.
FTR, we actually tried running first aid yesterday, but during or after the check the computer crashed. This might be an indication that there's indeed a problem with the file system, or it might be a red herring.
@wolfs ran the first aid again, this time it was successful. The error messages keep popping up, though.
He will try to upgrade to Catalina to see if it makes the problems go away.
Let's ask the community on Slack if they get the same message!
@wolfs just checking: you still see these, right?
Yeah, I didn't do the update to Catalina, yet.
Any plans for that?
Yep, still on my list of things to do.