Re: mv implemented as cp?!?!?!!? - problem!!

["Mark Paulus" <commpg at yahoo dot com>]

Unfortunately, it looks like it's a very tight timing issue.  If I put a 
system ("ls -ld foo tarballname"); inbetween the rename and rmdir
statement, depending upon the system state, it may or may not be
just enough of a delay to cause the script to work or fail.  Which
is why I think it's an atomic function issue, and debugging is very hard.
Kind of like finding a NULL pointer problem.  The very act of 
putting printf debug statements into the code is usually enough to
move the problem around, and make it hard as heck to find.


On Thu, 13 Sep 2001 12:36:05 -0400, Larry Hall \(RFK Partners, Inc\) wrote:

>Hi Mark,
>
>You're best bet for finding a solution to this problem is to debug Perl.
>This will give you the details you need to know how to proceed with a 
>proper solution.  Certainly it is Perl that is controlling the behavior
>in this case.  It may be possible to fix or override that behavior, but
>you'll need to find the right place to make it happen.
>
>Larry Hall                              lhall@rfk.com
>RFK Partners, Inc.                      http://www.rfk.com
>118 Washington Street                   (508) 893-9779 - RFK Office
>Holliston, MA 01746                     (508) 893-9889 - FAX
>
>
>
>At 12:26 PM 9/13/2001, Mark Paulus wrote:
>>This brings up an interesting issue.  I'm not sure where the problem
>>lies, or who is responsible for which piece but I will try to describe this
>>as cohesively and clearly as I can.  
>>
>>On my win2K box using FAT32, I have a perl script that does the following:
>>create a temp directory - foo
>>does a tar extract of  tarballname into foo/tarballname
>>rename foo/tarballname tarballname    
>>rmdir foo
>>
>>Now, it appears that if the directory structure under tarballname is
>>sufficiently large, the rename command is not actually completed 
>>before control is returned to the perl script, and rmdir foo is invoked.
>>This causes rmdir foo to fail with a Permission Denied error.  However,
>>after the script runs, foo is empty.  And if I place a sleep(1) between
>>the rename foo... and rmdir foo, the script completes successfully.
>>
>>Thinking about this as I write it, it seems to me that whatever is happening
>>in the rename() perl command, it is not being handled as an atomic unit,
>>and control is returning to the perl script before the system has actually
>>completed the rename operation.
>>
>>Is there some system switch that controls when/how control is returned
>>in these types of scenarios, as adding random sleeps is neither precise
>>enough, nor is it even preferred?
>>
>>
>>
>>On Wed, 12 Sep 2001 21:24:39 -0400, Matthew O. Persico wrote:
>>
>> >"Gerrit P. Haase" wrote:
>> >> 
>> >> Matthew O. Persico schrieb am 2001-09-12, 18:52:
>> >> 
>> >> >I have a /usr/X11R6 directory that I wanted to get out of the way. So I
>> >> >renamed it like this:
>> >> >
>> >> >cd /usr
>> >> >mv X11R6 X11R6.XFree.4.1.0
>> >> >
>> >> >It is taking multiple minutes. Looking at it in Windows Explorer, I see
>> >> >both X11R6 and X11R6.XFree.4.1.0 in existance at the same time.
>> >> >
>> >> >Is mv implemented as cp and del?!?!. Why not just rename the file, a'la
>> >> >UNIX, or even Windows? Or am I really missing something?
>> >> 
>> >> It works as you would expect, but only if there is no file opened in the
>> >> dir you want to rename.
>> >> You can try this.
>> >> Create a dir and fill it up.
>> >> Make sure that there is no file opened and you're not looking in any
>> >> subdirectory with explorer or another shell window.
>> >> Then rename it.
>> >
>> >Wow. Who knew? Obviously not me. I have to see how UNIX (solaris 2.6 in
>> >partcular) behaves when I get back to work tomorrow.
>> >
>> >-- 
>> >Matthew O. Persico
>> >New York City - we're still here.
>> >
>> >--
>> >Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
>> >Bug reporting:         http://cygwin.com/bugs.html
>> >Documentation:         http://cygwin.com/docs.html
>> >FAQ:                   http://cygwin.com/faq/
>>
>>
>>
>>
>>--
>>Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
>>Bug reporting:         http://cygwin.com/bugs.html
>>Documentation:         http://cygwin.com/docs.html
>>FAQ:                   http://cygwin.com/faq/




--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Bug reporting:         http://cygwin.com/bugs.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/