On Wed, Aug 27, 2008 at 07:59:39PM -0400, James E Keenan wrote:
> In the attachment, I show several lines from this file coverage report on a 
> Perl 5 module in the Parrot distribution:
>
> http://thenceforward.net/parrot/coverage/configure-build/config-auto-pmc-pm.html
>
> http://tinyurl.com/5fovnc
>
> Two statements, lines 261-262, each of which is a simple 'push'.  But 
> unlike other 'push' statements in this file, I get *two* entries in the 
> 'stmt' column for each of these two lines.  In the first case, line 261 is 
> encountered 2444 times -- but that's reported twice.  In the second case, I 
> also get two statement reports for the same line -- but the second report 
> is 0.
>
> These two lines were refactored from another location in the same file 
> today.  When I ran Devel::Cover on that earlier version, the two lines 
> appeared normal, i.e., no double 'stmt' coverage.  But the double coverage 
> today -- and specifically the '0' second statement on line 262 -- caused 
> the overall statement coverage for config/auto/pmc.pm to drop from 100% to 
> 99.3%.
>
> I don't understand why I'm getting double reportage on these two lines.  
> Any ideas?
>
> (Note:  The explanation shouldn't have anything to do, fundamentally, with 
> the branch in line 262.  That branch had only 50% coverage in the 
> *previous* coverage run as well.)
>
> Thank you very much.

There does seem to be something rather strange going on here.  The results for
the unshift a few lines earlier are also duplicated.  But more stragely, the
statement counts on lines 109 and 124 seem somewhat excessive.

My best guess is that some coverage database merging has gone awry.  Does that
sound plausible?  If you delete all the coverage information and start from
scratch does that "solve" the problem?  Did the runs take place on different
machines?  Are the clocks accurate wherever the runs were executed?  Did you
get any other messages from Devel::Cover?  If the problem can be reproduced,
is there some way I could reproduce it myself?  Have I asked enough questions?

-- 
Paul Johnson - [EMAIL PROTECTED]
http://www.pjcj.net

Reply via email to