I did this a while back using gct and published the results to p5p I
have since let the web site die (perl-site.com). I just posted it at
http://www.randomcontent.net/test-perl/coverage/ again, but be warned it
is rather old. After posting this information the P5P group didn't seem
that interested it was less than 1% as interesting a the 3 argument open
discussion.
On Thu, 15 Feb 2001 [EMAIL PROTECTED] wrote:
> (To those CC'd, you all either maintain or have at some point written
> a Perl coverage tool or profiler).
>
>
> As part of the QA process we need to do alot of test coverage analysis
> and, to a lesser extent, performance profiling. Our existing tools
> (Devel::Coverage, Devel::DProf, Devel::SmallProf) are a start, but
> need alot of work. We need really solid, tested, documented libraries
> *and* tools to pull this off.
I fully agree, testing with out metrics is like testing in the dark, you
don't know where you are or how much farther you need to go. The good
side of testing in the dark is most companies don't want the right
amount of testing so not knowing how much to do is an asset ;)
> Now, it strikes me that a profiler is just a coverage tool that also
> records times, and a coverage tool is just a profiler that doesn't.
The coverage tools do things like loop analysis to make sure you did 0,
1 and many iterations. If statements like this:
if (foo<bar)
must have foo<bar and foo=bar and foo>bar for 100% coverage to make sure
the programmer didn't really need to write the statement: if (foo<=bar)
> As such, I'd like the bunch of you to put your heads together and
> share tips, tricks and problems encountered creating your creations to
> hopefully improve the whole state of the art. Hooking into the
> debugger is something of a Black Art and it could use some airing out.
It shouldn't be but it is.
>
> This *doesn't* mean that I'm advocating a single, monolithic
> profiler/coverage tool. Keep your seperate implementations (or merge
> them, doesn't matter).
>
> We don't need fancy features, the most important things to improve are
> 1) Documentation and 2) Reliability. All the tools are historically
> lacking on those fronts. I'm already doing a bit of work on
> Devel::Coverage.
The way I left this project was that I was rewriting gct in perl so that
it would be eaiser to use. I'd love to see this thread go farther as I
feel this takes software testing to a level that people think it is at
but rarely is (even in medical software).
