Can someone tell me if this problem is fixed in -current ?

# mount -o rdonly <something> /mnt

# mount -t union /tmp /mnt

# cd /mnt

# echo hi >>nonexistent-file

(properly creates nonexistent-file in upper (tmp) fs)

# echo hi >>file-in-lower-fs

(EROFS in 1.6.2/i386)

Per the man page, write access to existing file in lower fs should copy to
upper fs and complete successfully.  It shouldn't matter whether the lower
fs is readonly; that's a typical use for a union fs!

I would hope this has been detected and fixed already, but my pr query
turned up nothing and I found no obvious reference in the cvs logs.


On a related note, has there been any discussion of a union fs option where
shadow dirs and files are created with their original ownership and perms
and not new ones derived from umask?  At least when the union mount is set
up by root and not j. random user, I think that should be an option.  I've
read of other users using elaborate fstabs and union mounts to keep their
local changes separate from their distributions, and I am very tempted to
do the same thing, but I Do Not Like the sound of the upper fs copies being
transparently created with different ownership and permissions.  It sounds
like 100 security holes singing in four-part chorus.

-Chap