Re: testserver down?

From: Michael Maierhofer <michael.maierhofer ***AT***student.tugraz.at>
Date: Fri, 11 Nov 2016 18:52:13 +0100
Subject: Re: testserver down?
Newsgroups: tu-graz.lv.vt
Organization: Technische Universitaet Graz
Message-id: <o050gd$io8$1@news.tugraz.at>
User-agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101 Thunderbird/45.4.0
References: <o02aka$ec0$1@news.tugraz.at> <op.yqqcltf0fwimhv@frohike> <o03v78$fv3$1@news.tugraz.at> <o045bi$1kf$1@news.tugraz.at> <o0460n$mp8$1@news.tugraz.at> <o04bdt$6k9$1@news.tugraz.at>
On 11/11/16 12:52, Magdalena Hackenberger wrote:
If I understood it correctly in the lecture, the plan was/is a testing
system with immediate feedback. There are still unresolved bugs and
problems with the system though.

Anyway, in most KUs, there is no feedback system at all and the tasks
are still doable. It's an additional service and there are plenty of
hints on the expected functionality in the newsgroup.

Am 11.11.2016 um 11:20 schrieb Michael Maierhofer:
Am 11.11.2016 um 11:08 schrieb Leo Prikler:

The test server is not a server. It was pointed out already for the
first assignment, that the autolab system is bugged, which is why there
is a manual starting of the tests in the first place. Leaking the test
data would only lead to students using them to cheat. Your program
should work for all inputs, you can write tests yourself.

Greetings,
Leo

Why use an online "testing" system if there's no immediate feedback (see
SNP or Operating Systems for test systems with automatic feedback)? It
makes 0 sense to have a system that you have to trigger manually if it
would be possible to test every submission automatically. It just sucks
to wait for feedback and not even being sure when feedback will be
available (twice a day, or more, or sometimes even no feedback at all).
And regarding the tests, one would think that writing your own tests is
sufficient, but if the test system apparently keeps on changing the
expected output how do you know which output is actually correct?



Well that's true, but most other KUs have a specific expected output one can rely on. This is most definitely not the case here. Especially since there is no real definition of the expected behaviour of the program in task 1. And I'm not even talking about the misleading definitions in task 2 or 3.

--
Michael Maierhofer <michael.maierhofer***AT***student.tugraz.at>
Telematik