Re: [hackers] Add realpath utility

From: Brad Barden <>
Date: Tue, 17 Nov 2015 18:42:29 -0600

On Mon Nov 16, 2015 at 05:48:07PM +0000, Dimitris Papastamos wrote:
> Sure but sbase readlink -f also covers this. It uses realpath(3).

readlink and realpath do cover the exact same ground, but as discussed
on irc, suckless readlink was broken with respect to -e and -m flags. we
talked about having realpath be the utility for those options, and leave
readlink to read links.

realpath -m is unfortunately non-trivial, and I tried to make it too
easy in my previous attempt.

I started over with a different tact, and implemented a realpathm()
function. It behaves identically to realpath(), but will canonicalize
paths with non-existent components.

Having realpathm(), implementing both readlink and realpath utilities is
almost exactly the same, the only difference being default behavior.

So in the attached patch is a working readlink with -e and -m flags
restored. 'realpath' could be implemented with a two-line shell script
or whatever, just calling readlink with the -f flag. That's nothing new,
as you pointed out, but the other flags work now.

There's probably plenty to pick on in here, and please do. Like I said,
it's non-trivial...this is the simplest I was able to make it and still
work in all the edge cases I could find.

I hope this is useful.

Extra side notes...I got started on this whole thing because I was
already looking at suckless tar. My big annoyance with it is that it
won't untar a relative tarball (with a "./" entry, tar tries to
remove(".")). It also makes no effort to confine itself to cwd, so a
(malicious, stupid, or accidental) tarball with
"foobarhaha/../../../../../etc/shadow" or whatever could ruin my day.
realpathm() gives a pretty easy way to do rudimentary checking before
extracting. I'm going to do that in my local sbase tree, put
realpathm() in libutil, maybe make a mkdirp()-ish with "relative to"
argument/check, and patch tar with it.


Received on Wed Nov 18 2015 - 01:42:29 CET

This archive was generated by hypermail 2.3.0 : Wed Nov 18 2015 - 01:48:17 CET