Re: cvs problem

[Sascha Wildner <saw@xxxxxxxxx>]

Matthew Dillon wrote:

>     Heh.  Well, the problem is that the old API is trying to purge the
>     namecache for rename() operations.  I was marking the node as being
>     unresolved but then when a new directory is created it re-resolves it
>     and the rest of the namecache under that dir becomes valid again.
>
>     The only real way to fix this properly is for me to implement the new
>     namecache semantics for rename.
>
>     I can hack a fix to support the 'old' API, which I will commit soon,
>     but what is going to happen is that any process chdir'd *INTO* the
>     subdirectory being renamed will get lookup errors if it tries to '..'
>     back out of it.  It isn't a usual occurance for that situation to occur,
>     fortunately.

Yeah, the bug is gone now and I'm getting lotsa

Oct  5 18:08:02 zoot kernel: [diagnostic] cache_resolve: relinked CVS
Oct  5 18:08:02 zoot kernel: [diagnostic] nlookup: relookup CVS
Oct  5 18:08:02 zoot kernel: [diagnostic] cache_resolve: relinked Root
Oct  5 18:08:02 zoot kernel: [diagnostic] nlookup: relookup Root
Oct  5 18:08:02 zoot kernel: [diagnostic] cache_resolve: relinked Repository
Oct  5 18:08:02 zoot kernel: [diagnostic] nlookup: relookup Repository
Oct  5 18:08:02 zoot kernel: [diagnostic] cache_resolve: relinked Entries
Oct  5 18:08:02 zoot kernel: [diagnostic] nlookup: relookup Entries

messages on the console upon cvs checkout (only for these four files).

Thanks,
Sascha

--
http://yoyodyne.ath.cx