Nicolas Williams wrote:
> On Wed, Mar 05, 2008 at 06:09:15PM -0800, Alan M Wright wrote:
>> Nicolas Williams wrote:
>>> On Wed, Mar 05, 2008 at 03:09:22PM -0800, Alan M Wright wrote:
>>>> Nicolas Williams wrote:
>>>>> Should a CR be filed against mv(1) here?  Or an RFE for an option to it
>>>>> that obviates the identical file check, at least on case-insensitive
>>>>> file systems?
>>>> Fromm an mv perspective this is indistinguishable from
>>>> having multiple links to a file.
 >>>
>>> nit: not if the link count is 1 :)
 >>
>> So ... it would be unfortunate if the thing you wanted
>> to rename was a directory.
> 
> I'm not sure I follow.

Directories never have a link count of 1.

>  rename(2) can rename files and directories; so
> can mv(1) (which uses rename(2)).  Directories cannot be hardlinked the
> way files can be.

because the link count is used for another purpose.

> The point is that mv(1) has a check that assumes case sensitivity.
> 
> Dropping that check would change its semantics in a case-sensitive
> filesystem -- it would let you rename a file even when a hardlink exists
> with the new name, in which case the hardlink would be unlinked during
> the rename.  So it can't just be dropped; mv needs to become smarter
> (i.e., more complicated) or users have to deal.
 >
> The fact that a work-around exists is nice -- this is a P4 or a P5.

Feel free to raise the RFE but (my opinion) I don't think changing
mv(1) to allow this by default is a good idea.  What do you do on
mixed-mode file systems, would mv(1) take the case-sensitive or the
case-insensitive perpspective?  Even if you add a new flag, you
would still have caveats - perhaps better to create a new command.

Alan
_______________________________________________
storage-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/storage-discuss

Reply via email to