Re: Separate obj tree (really emacs and hard links)

[Doug Alcorn]

address@hidden (Gary Oberbrunner) writes:

> then, quoting from the xemacs doc:
> 
>  | ...This works if any component of the pathname (including a non-
>  | terminal component) is a symbolic link as well, but doesn't work 
>  | with hard links (nothing does).
> 
> That last stmt is not true.  You *can* check if two files are the
> same; just check the inode numbers with stat.  

Yea, I've done some poking around in buffer.c file.  It stores the
"truename" as part of the structure associated with each buffer.  When
finding a file when find-file-compare-truenames is true, it merely
determines the truename of the file in question then rips through the
list of buffers comparing the stored truename to the truename of the
file in question.

What I expect is that something like the inode number should be stored
in the "buffer object".  Then you could have a new variable called
find-file-compare-inodes.  The last time I brought this up on
comp.emacs.xemacs I got complaints from the WIN32 crowd that FAT
derivatives don't support inodes.  I guess that would have to be a
compile-time option...


> And sure enough, when I try it on GNU Emacs, it does work.  Here's
> the message emacs gives me:
> 
>  ~/src/build/irix/final/spark-interface.c and ~/src/spark-interface.c are the 
> same file
> 
> Looks like I won't be switching to xemacs any time soon!

And the war rages on... ;)

-- 
 (__)  Doug Alcorn (mailto:address@hidden http://www.lathi.net)
 oo /  Unix Hacker
 |_/   "It's too late for paradise"