[fuse-devel] hard link doesn't preserve inode numbers

[Valient G. <va...@gm...>]

I just got a bug report from a user that mutt doesn't work properly
from within a FUSE based filesystem.   The reason stated is that mutt
makes a hardlink of a file and then checks that the inode numbers of
the two files match in order to ensure there was no race condition. 
This fails in FUSE because the files get assigned different inode
numbers (even though stat will show that there are 2 links to the
file)..

Does the communication between libfuse and the FUSE kernel module
assume that each pathname has a unique inode?  I suppose that with
multiple filenames refering to the same inode, that libfuse might
always pass one or the other as the filename to the filesystem
callback functions..  But is it necessary to distinguish which
filename was used to access hard-linked files?

Valient