This is the mail archive of the cygwin mailing list for the Cygwin project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Solved partially by findutils 4.3 - RE: "inode changed", "replaced while being copied"


-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

According to Jan Schormann on 1/18/2006 6:48 AM:
> Hi,
> 
> OK, one down, one to go ;-)
> New cygcheck.out to show the current config.

Can you convince your mailer to attach text files with a MIME type of
text, rather than application/octet-stream?  It makes it easier for me to
read if the attachment displays inline in my mailreader, rather than me
having to save it off to disk and open an external editor.  [Aside - maybe
the help instructions should recommend naming the file cygcheck.txt, as
some mailers use the file extension to choose the MIME type, and .out
typically implies binary, as in a.out.]

> 
> This one is SOLVED with the (experimental) findutils 4.3:
> 
> ----
> find: //desdata1/divisions/frameworks/share/Tools/BrainTools changed
> during execution of find (old inode number -411813144, new
> inode number
> -457114904, filesystem type is system) [ref 1114]

Explainable, since findutils 4.3.0 does use a different traversal method
than 4.2.27 (gnulib fts instead of a hand-rolled version).  Odd, since
coreutils 5.93 and findutils 4.3.0 both use gnulib fts as their traversal
method.  Also a bug in findutils - inodes should be printed as unsigned
numbers.

> 
> ----
> ~$ echo blub > //desdata2/universal/Sourcecode/test.txt
> ~$ cp -f //desdata2/universal/Sourcecode/test.txt /cygdrive/f/tmp
> ~$ mv //desdata2/universal/Sourcecode/test.txt
> //desdata2/universal/Sourcecode/test.exe
> ~$ cp -f //desdata2/universal/Sourcecode/test.exe /cygdrive/f/tmp
> cp: skipping file `//desdata2/universal/Sourcecode/test.exe', as it was
> replaced while being copied
> ----
> 
> The exe file is not copied, by the way.
> 
> These are consistent when I install 1.5.19-2 etc.
> After downgrading to 1.5.18-1, I'm back to usual.

Are they present in cygwin 1.5.19-2/coreutils-5.3.0-9?  If so, it is
something that change in how cygwin performs stat() on network shares (and
there were changes made in that arena).  In which case, an strace may be
interesting to show why stat() is giving different inodes on the same
network share file when it has not been modified.

> I guess there's something strange about our network drive configuration,
> but the problem only shows up in the newer cygwin, not in e.g. Windows
> explorer. Should I try to find out which package exactly makes the
> difference by installing all permutations, or will that not work anyway?

No - it is either coreutils or cygwin1.dll, and I am putting my bet on
cygwin1.dll.

> 
> Any hints as to what I could explore?
> Could these two observation have any connection?

Yes, both findutils and coreutils mv were complaining of the same
phenomenon - stat() was changing the reported inode of a file that the
program was trying to visit.

Also, do you know what filesystem (NFS, Samba, etc.) is on
//desdata1/divisions/ and //desdata2/universal/?  This will probably be
relevant, but Corinna will have to chime in here (as I cannot reproduce
the problem).

- --
Life is short - so eat dessert first!

Eric Blake             ebb9@byu.net
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.1 (Cygwin)
Comment: Public key at home.comcast.net/~ericblake/eblake.gpg
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFDzk0884KuGfSFAYARAmw+AJ9MamDdD1SLuzdKcHhWFVSSExKcmQCgzuyO
IGez1DbbZ1gox1/J1HVuf34=
=XbW6
-----END PGP SIGNATURE-----

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


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]