----- Original Message -----
From: "David Oswald"
If you have a look at the test reports linked to here:
http://static.cpantesters.org/distro/I/Inline-CPP.html
You'll see that suddenly Inline::CPP is failing with the majority of
smoke testers. Perplexing is that the first test to fail doesn't even
use Inline::CPP; it's just a "BEGIN{ use_ok( 'Parse::RecDescent' ); }"
test.
I couldn't find a report where that was the case.
I only looked at 3 reports:
http://www.cpantesters.org/cpan/report/8762a3e4-44e3-11e1-9d6f-f6dbfa7543f5
http://www.cpantesters.org/cpan/report/229dccd6-44de-11e1-9d6f-f6dbfa7543f5
http://www.cpantesters.org/cpan/report/e63683a4-44d9-11e1-bd28-ed17c154781f
and t/00load_prereqs.t passed fine in those reports - PERL5LIB and @INC were
still ok.
I think it's time this was raised on the cpan-testers list.
Maybe someone there has an idea as to why @INC is being altered for so many
of them when the Inline::CPP tests are run.
Could it be anything to do with the fact that Parse::RecDescent can't be
loaded with the -M switch ? On Windows I get:
##################
C:\_32>perl -MParse::RecDescent -e "print \"ok\""
Usage: perl -MLocalTest - <grammarfile> <classname>
C:\_32>
##################
And I wonder why it suddenly got worse. Have a whole lot of testers suddenly
updated something ?
David, can you find a tester/smoker there for whom 004 passed but 005 fails
? If so, we could then compare the 2 reports to find out what changed in his
setup.
It looks like for now I'm going to have to focus on documenting the
install process (which will state it may be necessary to pre-install
Parse::RecDescent before moving on to installing Inline::CPP).
It wouldn't do any harm - but isn't that what users do, anyway ?
Even if I use CPAN.pm to do the installation for me on a perl that doesn't
already have P::RD (and Inline), it will build and install P::RD before it
runs the Inline::CPP test suite. It's only these smokers that aren't
actually doing the 'make install' step (- or so it appears to me, at least).
Therefore, an instruction such as the one you're contemplating has relevance
only to cpan-testers ... and they're not going to take any heed of it,
anyway.
Cheers,
Rob
