17:38:23
scymtym
shinmera: i see the file-notify fault as well
17:39:12
scymtym
unless there has been some crazy cffi change, this seems like a case of "how could this work at all"
17:40:34
scymtym
the problem is that cffi treats the event filename as a pointer to a string although it is supposed to be a chunk of bytes in the event
17:41:03
shinmera
I do have cffi from master.
17:41:25
shinmera
Hmm, ok I'm getting the fault now.
17:41:28
scymtym
i.e. an event for a file named "foo" ends up as (cffi:mem-ref <sap #x666f5f> ...)
17:42:14
scymtym
it works if you replace (cffi:mem-ref (cffi:foreign-slot-pointer event '(:struct event) 'name) :string) with (cffi:translate-from-foreign (cffi:foreign-slot-value event '(:struct event) 'name) (cffi::parse-type :string))
17:42:58
scymtym
(i don't know the proper way to do this in cffi)
17:47:57
shinmera
Pushed a fix in f12dc2f
17:53:14
scymtym
great. though i still don't understand how it could work at all before this change
17:53:28
shinmera
Me neither. But it did somehow?
17:53:39
shinmera
I know I definitely tested it.
17:54:05
shinmera
And trial/kandria makes use of it to watch for asset changes.
17:54:27
shinmera
Though for that case I only watch specific files and not directories? Maybe that made a difference, dunno.
17:56:43
scymtym
would make sense since events pertaining to the watched path itself don't seem to contain a filename