[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]


Igor Khavkine <i_khavki@alcor.concordia.ca> writes:

> Would you really want that? I may want to keep passive translators
> on /mnt/c and /mnt/d which are FAT drives. But I don't use them very
> often, and I don't want locate/slocate polluting its database
> with filenames from those drives.

I dunno. Imagine that my setup included the following partitions

/home           (user playground)
/linux          (Linux /)
/mess/c         (M$ C: drive)

Everything except root is a passive translator. I wouldn't want the M$
stuff in my locate database ... and I'm not sure I want Linux things
in there, but I certainly want the /home tree.

The updatedb job will run in the wee hours, and if the translator for
/home times out properly, the results will depend on chance: if no-one
was logged in at this time, nothing from /home will end up in the

Essentially, we have a "interresting for locatedb" property that IMO
is completely orthogonal to a "mounted while updatedb runs" property.

Roland suggested fstab. Another possibility (if one wants more fine
grained control - e.g. one fs should show up in df, but not with
locate) is to explicitly give a list of filesystems to updatedb.

These are easily implemented, but limit the ability to mark a fs
interesting to root; which is IMHO not the Hurd spirit. Neal's
idea of translators registering themselves sounds better.

Summary: If you want a nice little Hurdish project, go for the dynamic
solution. If you just want locate to work, go for the former.

> If you have an automounter in linux you have the same effect, [...]

True, but as I understand it, "automounting" is the thing to do under
the Hurd, not an exception. I consider the above setup (nothing
mounted explicitly) quite normal.


Attachment: signature.ng
Description: PGP signature

Reply to: