Web lists-archives.com

Re: RFE: find <path> -d -size 0 => doesn't find empty directories




Greetings, Brian Inglis!

> On 2018-11-01 10:12, Andrey Repin wrote:
>> L A Walsh wrote:
>>> Unfortunately, due to directories really not being in the user
>>> disk data space, but in the MFT(zone) (I think), the size
>>> comes back as zero ('0') for directories.

>>> Would it be possible (if not problematic) for the cygwin
>>> emulation layer to return some non-zero value if the
>>> directory has actual entries in it (ignoring structural
>>> values like "." and "..")?  Maybe return as 'size' either
>>> a dummy number proportional to #entries (like 10*#entries),
>>> or something like summing up actual number (+1) of characters
>>> in the file list?

>>> Would that be difficult to do, or add?

>> Having something to this extent would be useful in case of searching for
>> directories with too many files, for example.

>> I'd vote for something like (entries << 7), which is closer to an average ext2
>> counter. No need to ignore anything.

> I believe readdir(3) overhead is already high, and adding extraneous lookups to
> add metadata which is not readily available under NTFS/exFAT would slow it even
> further.
> Do you really want readdir(3) or stat(3) to recurse to sum the entry sizes for
> each subdirectory?

If a number of directory entires is not readily available, nothing can be
done.
No recursion was intended.

> Some of us have some large messy directories more reminiscent of Unix systems
> than typical of Windows systems.

> $ time du -sh /tmp/
> 91M     /tmp/

> real    0m5.125s
> user    0m0.125s
> sys     0m1.077s
> $ time du -sh /var/log/
> 496M    /var/log/

> real    0m42.725s
> user    0m0.687s
> sys     0m9.139s



-- 
With best regards,
Andrey Repin
Friday, November 2, 2018 17:02:01

Sorry for my terrible english...


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple