| 1 |
On 06/27/2011 07:58 PM, Matt Turner wrote: |
| 2 |
> There seems to be an implicit assumption that the current code has |
| 3 |
> some kind of working test cases. :) |
| 4 |
|
| 5 |
I'm aware that I'm asking for test cases in context that seems to lack |
| 6 |
proper testing. |
| 7 |
|
| 8 |
|
| 9 |
> This is certainly not the case. Let me be clear, mistakes in the |
| 10 |
> current code come from having the same CFLAG, CHOST, etc strings |
| 11 |
> duplicated in many places. Refactoring the code would allow us to |
| 12 |
> catch mistakes like |
| 13 |
> http://git.overlays.gentoo.org/gitweb/?p=proj/catalyst.git;a=commit;h=db4323146ce27362948de6eab57e1dbe28240bde |
| 14 |
> much more quickly. |
| 15 |
> |
| 16 |
> It seems to me that test coverage would be much simpler if the classes |
| 17 |
> were refactored, since various combinations would use nearly identical |
| 18 |
> code paths. |
| 19 |
|
| 20 |
It would make some code pathes being taken more often but still leave |
| 21 |
the "leafes" ontouched without a test for each leaf. Right? |
| 22 |
|
| 23 |
What could work though is a throw-away test for refactoring only, say |
| 24 |
writing a piece of code making a text file listing all combination of |
| 25 |
CFLAGS offered from targets. If after the refactoring you get the very |
| 26 |
same text file out, that's a good indicator. Is the idea clear? |
| 27 |
|
| 28 |
Best, |
| 29 |
|
| 30 |
|
| 31 |
|
| 32 |
Sebastian |
Archives