droundy at jdj5.mit.edu
Fri Oct 10 12:27:42 UTC 2003
On Thu, Oct 09, 2003 at 08:07:50PM +0200, Luna Kid wrote:
> > >> argues that symlinks and file permissions are not "source", and are
> > >> the business of the build system rather than the RCS.
> Not knowing the full reasoning, I hope this is not the
> mainstream opinion...
As I said, I'm open to supporting symlinks, but not very interested in
working on them, largely because I'm not sure what the point of using
symlinks would be. Also, since I'm unlikely to use symlinks myself, I
don't want to add a feature that I don't know will be used.
> (Also, I guess relative (repo-internal) and absolute links
> must be treated differently, definitely keeping the linkage
> between repo-files, and probably taking the value of absolute
> links (links to external stuff) as plain text by default.)
At first I thought that repo-internal symlinks would be cool, but am now of
the opinion that all symlinks should be simply viewed as a special file
containing a string, with that string probably being restricted to not
contain a newline. Trying to track renames of the symlinked file seems
like wasted effort. What do you do when the file is removed? I presume you
leave the symlink there, and it's either a dangling symlink or a link to a
non-version-controlled file--which is more likely the case.
As far as possible, darcs commands should behave in the same way as native
operating system commands do. Anything else is likely to confuse users.
ln -s foo link
mv foo bar
mv bas foo
link points to foo.
More information about the darcs-users