[Python-Dev] cpython (merge 3.2 -> default): MERGE: Better test for Issue #15402: Add a sizeof method to struct.Struct (original) (raw)

martin at v.loewis.de martin at v.loewis.de
Mon Jul 23 23:44:47 CEST 2012

Zitat von Serhiy Storchaka <storchaka at gmail.com>:

On 23.07.12 22:46, Michael Foord wrote:

I am now prepared a set of 14 sizeof patches (should it be one issue or 14 individual issues in bugtracker?), and I feel a great desire not to write tests at all.

Without tests how can you have any confidence the patches are correct (or will continue to be correct)? Tests may not provide this. If we add a new dynamically allocated data or change the size of the old and forgot to reflect this in sizeof_, then non modified tests not notice this. _sizeof returns 42. What is 42?

42 is most likely not the right answer, as the size should be a multiple of four.

The point of writing tests for the sizeof code is that you detect your own mistakes in writing the test case. The value of the test case is not so much that it provides detection of regressions, but that you notice the bug once you write the test case.

You may argue that then there is no value in committing the test case, but there is also no harm in doing so.

In addition, it may catch regressions: if somebody changes the layout of an object, the corresponding test most likely fails. Maybe the fix is trivial and just in the test, but maybe the layout changed fundamentally, and the author forgot to change sizeof along with the structure change.

You wouldn't have to write 14 patches if implementing sizeof correctly was easy.

Please see the existing extensive tests for inspiration on how to add new ones.

Regards, Martin

More information about the Python-Dev mailing list