Re: [Gnu-arch-users] more on the merge-fest

[Samuel A. Falvo II]

On Tuesday 25 November 2003 11:55 am, Florian Weimer wrote:
> In an ideal world, you write test cases which solely rely on
> externally visible behavior.  In tla's case, you would create all test
> input data from scratch, using tla.  It is hard, often practically
> impossible to write a test suite that way, especially if you want to
> be able to actually run it. 8-)
>
> I agree that the usual approach is to add a test case which shows the
> behavioral difference you want to achieve, and adjust existing test
> cases accordingly (if they rely in internal details are normally
> considered hidden).

I would like to chip in briefly here: THANK YOU for using unit tests.  I 
failed to mention my appreciation for this earlier, when I first 
installed tla-1.1p8 (at least that's what I think it is).

I wish more software shipped with unit tests, and more package 
administrators had intestinal fortitude to require them, abide by them, 
and make sure software doesn't get shipped unless all unit tests pass 
100%.

You may or may not like the process of extreme programming.  But XP's 
test-first-by-intention (now more formally called "test-driven design") 
is quite valuable, even in isolation of the rest of the XP process.  
Since having adopted that method of development, I rarely ever use gdb 
for anything anymore.

--
Samuel A. Falvo II