svn commit: r291929 - head/tools/tools/nanobsd

[Ian Lepore]

On Mon, 2015-12-07 at 11:07 -0800, Garrett Cooper wrote:
> > On Dec 7, 2015, at 10:13, Ian Lepore <ian at freebsd.org> wrote:
> > 
> > > On Mon, 2015-12-07 at 10:24 +0000, Warner Losh wrote:
> > > Author: imp
> > > Date: Mon Dec  7 10:24:38 2015
> > > New Revision: 291929
> > > URL: https://svnweb.freebsd.org/changeset/base/291929
> > > 
> > > Log:
> > >  Make sure to quote the arg after -n and -z tests.
> > 
> > While I am generally a quoting fanatic and would much rather
> > overquote
> > than underquote in shell scripts, fyi it's not actually needed for
> > the 
> > -n and -z tests.  test(1) and shells seem to be smart enough to see
> > the
> > ']' (which is just an arg to test, not shell language syntax) and
> > know
> > there isn't a string in front of it.
> 
> Warner's commit was good. There's a more bulletproof way to do this
> to make it more portable (in most cases, not all), but it would be
> overkill:
> 
> [ "x$x" != x ]
> 
> Also, his commit protects against variables with spaces in them doing
> unexpected things with test(1).
> 
> Thanks,
> -NGie


Arrrgggghhhh!  No no no no, a thousand times NO.

There is NO EXCUSE for that ugly unnecessary xfoo stuff.  Quoting
properly supplies all the bulletproofing needed for string comparisons.

It does look like I was wrong about all shells/test implementations
handling the closing bracket for -n and -z as an empty-string indicator
though, so quoting is a good solution for that.  (Quoting is a good
solution for everything, it drives me nuts that I can't use spaces in
directory and filenames because so many unix scripts fall on their
faces with parameter passing and other tricky quoting situations).

-- Ian