Crash when running in alpine podman container

[salatfreak]

Running the parcel development server in an alpine podman container crashes because of an error being thrown in @parcel/watcher. The error is caused by a signal interrupting the poll system call in the linux inotify backend, that is sent by the __membarrier function in the musl libc implementation to synchronize threads when it fails to make a MEMBARRIER_CMD_PRIVATE_EXPEDITED membarrier syscall.

I have published a minimal reproduction of this kind of error with some further description of the problem in this repository.

The problem could be fixed by either masking signal 34 while polling or by recovering from EINTR errors instead of crashing. Masking the signal could be accomplished by replacing poll with ppoll and using the sigmask argument.