i just read with interest the recent thread about release being
asynchronous. i'd just worked that out for myself, the hard way. it might
be worth mentioning that in /usr/include/fuse/fuse.h.
my question, though, is whether i can do anything about later operations
overtaking it. say someone's running a script against my file system, and
they open, write, close, and unlink, all on the same pathname, all from a
single thread. my problem is that very rarely my file system sees the
unlink before i see the close. is this a bug that i should put together a
proper test case for, or a well-known feature that i need to work around?
and if the latter, what are the usual recommended work-arounds?
Elliott Hughes, http://www.jessies.org/~enh/