`IN_Q_OVERFLOW` arrives with watch descriptor -1, so the path lookup
failed and the event was silently discarded. It is now delivered
with an empty path so consumers can rescan after the kernel dropped
events.
The limits were only written back after a successful body, so a
failing limit test left the shared kernel at the lowered values and
every later run failed with ENOSPC. The helper now restores in a
`defer` and can lower a chosen subset of the limits.
Events were buffered with `bufferingNewest(512)`, so a burst of
changes silently lost all but the newest 512 events whenever the
consumer lagged. The stream now buffers without limit, and
`init(bufferingPolicy:)` lets callers choose a bounded policy.
`deinit` closed the inotify descriptor while its dispatch source was
still active. The kernel drops the epoll registration on close, but
libdispatch keeps its own; an instance created afterwards that reuses
the descriptor number could inherit that stale state and never
receive events. Roughly one test run in three lost a single event
this way.
The reader is now cancelled in `deinit` and the descriptor closed in
its cancel handler, as libdispatch requires.
inotify exposes a /proc interface to limit kernel memory usage. If those
limits are set too low, inotify cannot add all watches. The integration
test verifies, that Inotify yields an error in that case.
Each inotify instance produces events for paths in its watch list. Each
item in the watch list is identified by its watch descriptor. Different
paths can be watched for different events.
T. R. Bernstein