gnu-arch-users
[Top][All Lists]
Advanced

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

Re: [Gnu-arch-users] extended attributes


From: Brian May
Subject: Re: [Gnu-arch-users] extended attributes
Date: Mon, 05 Jan 2004 16:35:09 +1100
User-agent: Gnus/5.1002 (Gnus v5.10.2) Emacs/21.3 (gnu/linux)

>>>>> "Aaron" == Aaron Bentley <address@hidden> writes:

    Aaron> For myself, I'm not sure of this.  While the link-breaking
    Aaron> behavior is useful to Arch, I'm tempted to call it a bug.
    Aaron> My expectation of an editor is that it doesn't change the
    Aaron> properties of the file (e.g ctime, permissions, # of links)
    Aaron> -- it just changes its contents (and atime/mtime).

I tend to agree; generally speaking, the write-and-rename procedure
attempts to create an identical file, but sometimes it isn't always
possible to create an identical file with the same
{uid,gid,permissions}.

eg. if you are editing a file that is not owned by you. Even worse, if
you want to update a file (that you have write access to) that lives
in a directory that you don't have write access to. You can't do it!
(disclaimer: I haven't tested either situation recently).

Eventually editors will understand extended attributes though, so it
should be possible for them to update the extended attributes. I don't
see any solutions for preserving file permissions though.

The reason I see for this "broken" behaviour is you can share your
"work tree" with the "original tree"; even this has its issues, you
only need one program to overwrite the file instead of replacing it,
and the original tree is invalid. Last I checked, autoconf and
automake, etc were guilty of doing this, not sure if the situation has
changed.

I suspect this will remain a controversial issue.
-- 
Brian May <address@hidden>




reply via email to

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