Return Styles: Pseud0ch, Terminal, Valhalla, NES, Geocities, Blue Moon.

Pages: 1-

What is a "good" code coverage %?

Name: Anonymous 2010-07-20 6:37

What is the scale for what is considered "good" in terms of code coverage?

Are we talking 70%? 80%? 95%?

Name: Anonymous 2010-07-20 6:44

More than 50% is already near enterprise level.

Name: Anonymous 2010-07-20 8:38

>>2
Are you from Microsoft?

Name: Anonymous 2010-07-20 9:52

Fuck you on? All code should be code, i.e. 100%

Name: Anonymous 2010-07-20 12:40

>>4
has never written anything with more than 100 LOC.

Name: Anonymous 2010-07-20 12:51

>>5
18 more lines; before #190  2 seconds ago                     328,0-1       Bot

Name: Anonymous 2010-07-20 13:27

Delightful buzzword salad.

Name: Anonymous 2010-07-20 17:13

I don't even know what code coverage means.
The code-documentation ratio?

Name: Anonymous 2010-07-21 1:25

Code coverage means that you're testing all possible paths for the code to take.
 I would focus more on obscure paths. Sometimes it's not about trying to get everything but uncovering strange paths that can cause issues.

Name: Anonymous 2010-07-21 3:15

>>9
I think this is why we need provers.

Name: Anonymous 2010-07-21 4:45

>>10
This. Testing only confirms the presence of bugs, it doesn't prove their absence. If your code can't be proved to be correct, something is already wrong.

Name: Anonymous 2010-07-21 7:21

largely depends on how leveraged your workflows are, but I say around 0.80 is the optimum ratio

Name: Anonymous 2010-07-21 8:14

>>11
If your code can't be proved to be correct, something is already wrong.
Virtually every piece of software ever written is wrong? Most (read: all except one or two academic) languages aren't specified fully enough for a formal proof to even be possible.

Name: Anonymous 2010-07-21 8:44

>>13
Virtually every piece of software ever written is wrong?
Correct. Show me a bugtracker with no entries and I'll show you software even the author doesn't use.

Name: Anonymous 2010-07-22 1:50

>>13

"Computer checked proofs of program correctness are now possible for pure LISP"

-John McCarthy, 1980

Name: Anonymous 2010-07-22 1:59

>>14
Or a bugtracker that's not used.

Name: Anonymous 2010-07-22 8:06

Name: Anonymous 2010-07-22 8:36

>>17
Naggum my ANUS!

Name: Anonymous 2010-07-22 14:36

>>17
See this is where it gets way off track. When I mentioned provers, I was talking about the coverage angle. If your spec is wrong, if your unit tests are wrong or if you're simply off your meds that day and regression testing last month's code then it's all the same: you've fucked up somewhere and the results are nigh worthless.

The utility of the prover with respect to coverage is that you end up having to write/perform a lot fewer tests.

another assumption is that specifications need to be as precise as the program is.  this is not so.  code contains a lot of redundancy of expression and information and little expression of intent and purpose.  specifications should contain what the code does not.
With respect to this discussion (ca. my post), this is the best passage from that link. A simple specified requirement can set the prover about quite a large amount of code, if it is all involved in producing the result. If that result is the only requirement of that code, a very simple specification can be used to prove a (relatively) very complex portion of the program, and more importantly it gives you coverage even if the spec is wrong. You will know your components work as understood according to specifications even if not as intended by the client. (This might seem useless, but imagine it isn't true: now it's wrong and broken.)

Name: Anonymous 2010-07-22 17:00

>>17-18
*African-Amerigum

Name: Anonymous 2011-01-31 20:46

<-- check em dubz

Name: Anonymous 2011-02-03 5:05

<

Name: Anonymous 2011-02-04 18:31


Don't change these.
Name: Email:
Entire Thread Thread List