A few months ago I took over Inline::CPP and have been working to improve its smoke test ratio. As you might guess marrying C++ with Perl via Inline has been fraught with challenges. When I started Inline::CPP was passing about 8% of its smoke tests (v0.25). As of the current mainstream version (v0.33) it's passing about 75% of smoke tests.
However, my development branch has suddenly taken a turn for the worse, and in an area that shouldn't be an issue. If you look at the static test summary page: http://static.cpantesters.org/distro/I/Inline-CPP.html you will see that v0.33_004 and 0.33_005 (especially the latter) have begun failing miserably. The primary issue is illustrated in the following report (it's easy to find a good example, as there are many similar reports): http://www.cpantesters.org/cpan/report/8e352ea2-488b-11e1-9d6f-f6dbfa7543f5 Looking over the report you will see that test 00load_prereqs.t passes ok. That test uses Test::More to do a "use_ok( 'Parse::RecDescent' ). However, the next test: 01basic.t fails because it cannot find the dependency Parse::RecDescent. The 00load_prereqs.t test also spits out a "diag()" message that shows the PERL5LIB environment variable so that I can confirm that the paths look reasonable. But as I mentioned, the use_ok() test passes, while the 01basic.t test (which invokes Inline::CPP, that in turn requires Parse::RecDescent) fails. What is particularly frustrating is that between Inline::CPP v0.33 and v0.33_005 there hasn't been any change to the portion of the code that invokes Parse::RecDescent. And Parse::RecDescent is listed in the PREREQ_PM field by Makefile.PL for ExtUtils::MakeMaker. Yet over the course of a few weeks suddenly the tests have all begun to fail (or the vast majority) where they were once passing. I've been soliciting assistance from the wise folks over at inl...@perl.org, where Rob (sysiphus, maintainer of Inline and Inline::C) suggested that possibly a recent upgrade to the smokers has caused this problem for me. I'm also theorizing that CPANPLUS may be at fault, though I can't figure out how so. If there is anyone still reading, and interested in helping me figure this out please let me know. Here's a link to the current dev. distribution of Inline::CPP on CPAN: http://search.cpan.org/~davido/Inline-CPP-0.33_005/ So the questions: What may have changed with a whole bunch of smokers all within a few weeks of each other causing Inline::CPP to suddenly start failing the majority of the smoke tests in this "missing dependency" way? Could it be some configuration common to the smoke testers? Could it instead be a recent change to CPANPLUS? Where do I go from here in resolving the problem? Thanks for your time, Dave -- David Oswald daosw...@gmail.com dav...@cpan.org davido on PerlMonks
