Making a test suite for uniq.

[Manuel Novoa III]

Hi Rob,

On Wed, Sep 14, 2005 at 09:51:47AM -0500, Rob Landley wrote:
> So I just checked in a new test suite for uniq (trying to get full susv3 
> coverage), and I've found two failure cases.  Both of these are explicit 
> violations of the spec, which gnu gets right.
> 
> 1) In gnu, "uniq -c" has a space between the number and the line, and we have 
> a tab.  SUSv3 actually specifies that this should be a space. 
> 
> OUTPUT FILES
>   If the -c option is specified, the output file shall be empty or each line 
>   shall be of the form:
>   "%d %s", <number of duplicates>, <line>
> 
>   otherwise, the output file shall be empty or each line shall be of the form:
>   "%s", <line>
> 
> (Technically the leading space is a violation of the spec too, but since gnu 
> does it I phrased the test to accept but not require it.)

Older versions of gnu uniq also seperate also seperate the count from
the line with a tab rather than a space.

> By the way, I fixed this in my checkin last week, and you reverted it along 
> with everything else.  I also note that the warnings you put back are still 
> there.

I said I'd get to it this week.  The week isn't over yet.

> 2) In gnu, "uniq -d -u" prints nothing at all, and this follows the spec:
> 
> -d  Suppress the writing of lines that are not repeated in the input.
> -u  Suppress the writing of lines that are repeated in the input.
> 
> Therefore, you specify both you suppress both.

Again, older versions of gnu uniq didn't behave this way.  They treated
-d and -u as mutually exclusive and the last one was honered.

> 
> Rob
> 
> P.S. Yeah, I get a bit pedantic when I'm claiming compliance with an official 
> formal specification.  But currently, we're wrong.

Fine with me.  I'll incorporate those changes.  I'll also add an option
to the config system to toggle gnu-ish leading whitespace if you like.

Manuel

P.S.  Older version tested was uniq (coreutils) 4.5.3