bug-coreutils
[Top][All Lists]
Advanced

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

Re: bug in rsnapshot 1.2.1 due to change in /bin/cp 5.93


From: Tim Ruddick
Subject: Re: bug in rsnapshot 1.2.1 due to change in /bin/cp 5.93
Date: Thu, 6 Apr 2006 21:52:07 -0700
User-agent: Mutt/1.4.2.1i

Eric--

Thank you for your swift and gracious reply, especially since I hadn't
checked the mailing list archives to see whether the bug had already
been noted.  My bad.

Whatever way rsnapshot decides to resolve this, I have a workaround that
will work fine for me indefinitely (using rsync instead of /bin/cp).

Cheers!
Tim.

On Fri, Apr 07, 2006 at 02:22:15AM +0000, Eric Blake wrote:
> > FC5 uses /bin/cp version 5.93, and this version doesn't like a trailing
> > slash on the destination directory name.  /bin/cp v5.2.1 (the version on
> > Fedora Core 3) used to accept it.  Looking at the man page, it appears
> > version 5.93 can be made to use the old behavior by passing '-T' on the
> > command line.  Vexingly, v5.2.1 doesn't like the -T option.  Examples:
> 
> Thanks for the bug report.  However, this issue has already come up in
> the past:
> 
> http://lists.gnu.org/archive/html/bug-coreutils/2006-02/msg00069.html
> http://lists.gnu.org/archive/html/bug-coreutils/2005-11/msg00285.html
> 
> The consensus seemed to be that POSIX is ambiguous on what is
> required, and that while cp should someday be fixed to either be
> consistent across all platforms or to honor the semantics of the
> underlying rename(2) syscall, no one was willing to delay 5.94
> for such a patch.  So your issue is still present in coreutils 5.94
> (the latest stable version).
> 
> All I can recommend now is that rsnapshot be patched to avoid
> this portability trap of cp, since other platforms (such as Solaris
> or BSD) have different rename semantics in this corner case.
> 
> -- 
> Eric Blake




reply via email to

[Prev in Thread] Current Thread [Next in Thread]