Problems with xattr / v2.1.6 Monterey (Intel)

[roemer]

OK so is there a difference to hfs/apfs in the current behavior? Cos we could make sure to remove empty xattrs. What creates them?

Actually, I think we are doing well:
Regarding normal xattr, zfs matches HFS/ZFS.
Regarding com.apple.FinderInfo, zfs with mimic=apfs|off matches APFS, and with mimic=hfs matches HFS.
Regarding com.apple.ResourceFork, zfs seems to match apfs and zfs, with the exception that apfs/hfs do not list empty resource forks as xattr.

Example with APFS below; note how it gives no error message when writing a ResourceFork, but then I cannot list it as extended attribute, though accessing it using ..namedfork/rsrc seems to work:

Code: Select all

touch testfile-apfs
xattr -w com.apple.ResourceFork ""  testfile-apfs
ls -l@ testfile-apfs
-rw-r--r--  1 user  staff  0 26 Feb 19:05 testfile-apfs
xattr -lx testfile-apfs
       [no output, but also no error]
cat testfile-apfs/..namedfork/rsrc
       [no output, but also no error]
wc testfile-apfs/..namedfork/rsrc
       0       0       0 testfile-apfs/..namedfork/rsrc

With ZFS 2.1.6-99, a 0-length com.apple.ResourceFork xattr is explicitly listed, otherwise everything is the same as with APFS:

Code: Select all

touch testfile-zfs
xattr -w com.apple.ResourceFork '' testfile-zfs
ls -l@ testfile-zfs
-rw-r--r--@ 1 user  staff  0 26 Feb 19:06 testfile-zfs
   com.apple.ResourceFork   0
xattr -lx testfile-zfs
com.apple.ResourceFork:
00000000
cat testfile-zfs/..namedfork/rsrc
       [no output, but also no error]
wc testfile-zfs/..namedfork/rsrc
       0       0       0 testfile-zfs/..namedfork/rsrc

I originally tried only the xattr access and saw a difference in output; but looking at the full picture above including using ..namedfork/rsrc , I think this should be fine. I actually like that ZFS shows with xattr that there is an empty resource fork, while APFS seems to hide it...

[lundman]

Awesome, thanks for going through all the testing. I'll work on a 1.2.7 release if the corruption gets fixed.

[roemer]

Great - many thanks for your fantastic work!
Very appreciated, and I am looking forward to the release version for Apple Silicon / Ventura

By the way - now that the com.apple.mimic setting has some new meaning, would it make sense to document those in the zfsprops man page,
as well as in the "zfs get" and "zfs set" help texts?

[o3x_prokyon]

Yes, thanks for your great work and testing roemer and lundman!!! Last week my new M2 Mac has arrived ! Are the problems with v2.1.6 also present on Apple Silicon? I guess they are, because the filesystem is the same... So I'm also looking forward to the official release of 2.1.7 for apple silicon/Ventura.