IMHO, there are two different things to consider:
- a non-regression tests suite which is only about regressions. It contains
tests that once proved to be a BUG which are fixed and we don't want to see
back again. It contains also complete and silly tests that are not expected
to fail (1.0 == 1 is such a silly test. Sometime very idiot tests fail
because of a very innocent change). My OpenLisp very BIG regression tests
contains tests like that which proves I'm unfortunately right. A
non-regression tests suite is not expected to fail. It only fails on change
that is a **regression** on what used to work.

- a missing features test that shows something is missing or buggy you add a
test case for it in bug-test

Suppose you think of new test like:
int a; double d;
a = d = 3.14;

If it works, it goes to non-regression test, if it fails it goes to bug test
and when fixed goes to the first one.

This way, non-regression test always works and you know what you must work
on with second test suite.


-----Original Message-----
From: Tinycc-devel []
On Behalf Of Michael Matz
Sent: jeudi 13 octobre 2016 14:49
Subject: Re: [Tinycc-devel] cleanups


On Mon, 10 Oct 2016, grischka wrote:

> In general, often when I see people adding tests I think:  "Well you 
> just fixed that, what's the point?  I'd rather see what's still broken."

The point of course is to not break things ever again, after somebody got to
the length of fixing something.


Tinycc-devel mailing list

Tinycc-devel mailing list

Reply via email to