Re: logical paths

[Kartik K. Agaram]

> ".." occurs in a lot of places.  Will those patches put a wrapper
> around every system call that uses a file name?  And what, exactly,
> will that wrapper do?
>
> Will the patch add a new option to every coreutils command that uses
> file names?  That sounds pretty intrusive.

What will the wrapper do? Hmm, I had a look at coreutils-5.97/ls.c for 
starters, and my plan had been to compute absolute_name in gobble_file (ie 
for every non-option argument to a coreutils command) using pwd if it's a 
relative path. I simply collapse away any number of '..'s with elements 
from pwd, and compute an absolute path before feeding it to stat. Yes, 
it'll call for an extra option, but I think it'll be consistent across all 
of coreutils and the code is mostly a single function to convert relative 
path to absolute path that we can use across all of coreutils. These are 
my tentative thoughts.

I understand that adding an option or environment variable is a major 
addition, so please bring up objections. In particular, can you point out 
some other places than non-option arguments that we need to handle '..'?

Thanks,
Kartik