develooper Front page | perl.perl5.porters | Postings from August 2013

[perl #119351] Perl 5.18.1 breaks Devel::Cover

Thread Next
James E Keenan
August 18, 2013 13:10
[perl #119351] Perl 5.18.1 breaks Devel::Cover
Message ID:
# New Ticket Created by  James E Keenan 
# Please include the string:  [perl #119351]
# in the subject line of all future correspondence about this issue. 
# <URL: >

At some point between Perl 5.18.0 and 5.18.1, Devel::Cover began to fail 
most of its tests.

Here are the CPAN Testers Perl/Platform version matrices for 
Devel::Cover 1.06 and those two versions of Perl:;perl=5.18.0;reports=1;perl=5.18.1;reports=1

With one unusual exception, D::C 1.06 has built and tested successfully 
on Perl 5.18.0.  In addition to the platforms listed there, yesterday I 
successfully built, tested and installed D::C 1.06 on Perl 5.18.0 on the 
same Darwin/PPC machine on which I have installed D::C for many years. 
(I had not yet upgraded to Perl 5.18.1 on that machine.)

However, when we turn to the matrix for Perl 5.18.1, we see D::C failing 
across the board.  I haven't inspected all the failure reports, but for 
those I have I observe:

* They are the same failures I got yesterday when trying to install D::C 
1.06 on top of a new Perl 5.18.1 on the same Linode (i686) on which I 
have installed both Perl and D::C for many years.

* They are the same failures seen in the one, unusual failure report for 
5.18.0 noted above 

When diagnosing this yesterday, I tried to install a slightly older 
version of D::C (1.02) on Perl 5.18.1 on the Linode.  I know that this 
version worked with Perl 5.18.0 because I had installed it there earlier 
(but I had to remove the 5.18.0 for other reasons).  D::C 1.02 failed on 
5.18.1 just as D::C 1.06 does.  So I inferred that the problem lies in 
changes to Perl rather than changes to Devel::Cover.

This has been reported in D-C's bug queue:

In that report, the poster notes what I also observed:  On top of 
5.18.1, none of the cover_db/ databases are being written to.  This 
would explain why, in the individual failure reports, the "Got" 
(left-hand) side of the output chart shows blanks where the "Expected" 
(right-hand) side of the chart shows coverage percentages.

Can someone with a big enough box try at a 'git bisect' on this?

Thank you very much.
Jim Keenan

Thread Next Perl Programming lists via nntp and http.
Comments to Ask Bjørn Hansen at | Group listing | About