Fun w/ test :)

[Denys Vlasenko]

Paul, hi. :)

On Monday 21 July 2008 20:35, Chris Larson wrote:
> clarson at foul busybox% ./busybox test -f =
> clarson at foul busybox% ./busybox test \( -f = \)
> test: closing paren expected
> 
> Hehe.  I highly doubt anyone would actually name a file =, but what the heck, figured you'd
> want to know about any cases that confuse the parser.  It's interesting to note that -f =
> works, but ( -f = ) does not.  No customer issue on this of course, just made myself try
> lots of combinations during the testing.
> 
> I found a case where the "unknown operand" error is seen.  If it sees an operand after an
> expression, which is not -o or -a, it gives that error:
> clarson at foul busybox% ./busybox test -f = -a
> clarson at foul busybox% ./busybox test -f = -a -e foo
> test: -e: unknown operand
> 
> Due to the precedence, that's seen as two expressions without a -a or -o to join them.
> The error message could be confusing in this case. -e is, of course, a valid
> operator, and -e foo is a valid expression, but -e is not valid -there-, only
> -a or -o are.


So, now, do you see why x"str" trick is actually not so silly?


On Thursday 15 May 2008 02:20, Paul Fox wrote:
>  > You're talking again of som general case.  In this particular case:
>  > 
>  >   [ x"$foo" != x ]
>  > 
>  > it's totally useless quoting an empty string after the 'x' on the
>  > right side of the condition.
> 
> 
> sigh.  why would you write this at all?  either say what you
> mean, and write:
>     [ "$foo" != "" ]
> 
> or say it in the usual shell idiom and write:
>     [ "$foo" ]
> or
>     [ -n "$foo" ]
> 
> but this
>     [ x"$foo" != x ]
> or this
>     [ x"$foo" != x"" ]
> 
> are just silly.
> 
> paul
> =---------------------
>  paul fox, pgf at brightstareng.com

--
vda