unionfs kqueue?
Hi, Does unionfs work with kqueue? When I run `tail -f` on a file residing on unionfs with cd9660 underneeth and md+ufs over it, it doesn't detect changes. The changes are immediately visible, just not with tail -f. signature.asc Description: OpenPGP digital signature
Re: unionfs kqueue?
Heya, Does unionfs work with kqueue? When I run `tail -f` on a file residing on unionfs with cd9660 underneeth and md+ufs over it, it doesn't detect changes. The changes are immediately visible, just not with tail -f. Hmm. When you start the tail -f, does the file reside on the cd9660 or already on the md? See if tail -F does a better job. My guess would be that, since you cannot modify a file on any filesystem except the top one, unionfs must change semantics of open so that even opening for writing or appending silently creates a new copy of the file on the top filesystem (if the file didn't reside there already). As tail -f still has the lower-layer file open, it never notices that there's a new file by the name. (this behavior is the same as echo foo foo tail -f foo # in another terminal echo bar bar mv bar foo which also fails to notice the new data) So Far, Karsten BearPerson Behrmann p.s.: this is probably why the -F option was added to tail -- Open source is not about suing someone who sells your software. It is about being able to walk behind him, grinning, and waving free CDs with the equivalent of what he is trying to sell. signature.asc Description: PGP signature
Re: unionfs kqueue?
Karsten Behrmann wrote: Heya, Does unionfs work with kqueue? When I run `tail -f` on a file residing on unionfs with cd9660 underneeth and md+ufs over it, it doesn't detect changes. The changes are immediately visible, just not with tail -f. Hmm. When you start the tail -f, does the file reside on the cd9660 or already on the md? The file resides entirely in the upper layer, in the md. See if tail -F does a better job. Will do. My guess would be that, since you cannot modify a file on any filesystem except the top one, unionfs must change semantics of open so that even opening for writing or appending silently creates a new copy of the file on the top filesystem (if the file didn't reside there already). Yes, AFAIK it does that. As tail -f still has the lower-layer file open, it never notices that there's a new file by the name. Not in my case - there's no such lower-layer file. signature.asc Description: OpenPGP digital signature