James Carlson wrote: > Richard Brown writes: >> Recently, the project team responsible for modifying the archive >> utilities (cpio, tar, etc.) found that the tests were failing when >> they tried to restore the SUNWattr_rw extended attribute file which >> contained the OPAQUE attribute. The project team realized that since >> OPAQUE was approved as an unmodifiable optional file attribute (see the >> spec for PSARC 2007/315, section 3.3 "Third-Party Requested >> Attributes"), then it should be part of the read-only view represented >> by the SUNWattr_ro extended attribute file. > > Isn't the same true of the "SETTABLE" attribute? It's listed as > failing with EPERM on all attempts, and in the SUNWattr_rw section, > just like OPAQUE. >
The SETTABLE attribute was removed with Rescind SETTABLE Attribute [PSARC/2007/444 Self Review] > Currently, it looks like the SUNWattr_ro items are all things that can > be determined at mount time (FSID and MDEV), and thus the fact that > these can't be changed is without consequence. But where does the > OPAQUE flag come from? How does it ever get set in the first place, > if it's a read-only attribute? > OPAQUE is a BSD specific attribute, which is used as part of unionfs. > Perhaps what's needed is a better way to tell backup utilities that > certain attributes shouldn't be part of the archive. > Everything in the readwrite view should be archived. -Mark
