On 2016-02-23 16:57, Jeroen Demeyer wrote:
> On 2016-02-23 15:50, Clemens Heuberger wrote:
>> Bottom line: I cannot reproduce 11 month old results anymore.
>
> Personally, I think this is a flaw in the review process for Sage tickets.
> Reviewers should check whether existing code might break with a patch, but
> that's not always done.
>
>> Of course, I could have opened a ticket and included my code into the
>> src/sage/tests directory. But I cannot imagine that this approach would scale
>> very well. I assume that there are many papers out there with a substantial
>> amount of CPU time required for reproducability.
>
> What if you just add a simple case? Imagine your code computes f(n) for all n
> up
> to 10^6. Then you could use f(10) as test case.
My impression after one day of debugging is that while we may have a very good
coverage of tests covering specialised methods, we have very few tests which
test behaviour of Sage as it is actually used.
If I counted correctly, there are tests covering 6 publications. Some of those
are rather isolated _examples_ for how to use sage.
There are also some tests in src/sage/tests which do not correspond to
publications at all.
So it seems that there are almost no tests which actually do solve some
mathematical problem of any sort.
This all boils down to the following question:
What is the policy concerning src/sage/tests?
Some subquestions coming to my mind:
* Shall tests related to publications be moved to another directory
(sage/tests/publications)?
* What files are welcome there?
* How strict are coding standards? Do we enforce PEP8 compliance?
* Shall there be a mechanism that the tests are not included in make ptestlong,
but are tested less regularly?
I cannot believe that our code was the only code which was broken over the last
11 months by 2 to 3 (depends on how you count my second and third problems)
different changes of behaviour without deprecation period.
--
You received this message because you are subscribed to the Google Groups
"sage-devel" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to sage-devel+unsubscr...@googlegroups.com.
To post to this group, send email to sage-devel@googlegroups.com.
Visit this group at https://groups.google.com/group/sage-devel.
For more options, visit https://groups.google.com/d/optout.
