This is a minor concern because most people won't be installing it due to the virtual defaulting to perl. Its just something we should look into doing as it looks more sensible than what we're currently doing. So I'm just sticking this here so It doesn't get forgotten. EUMM's own Makefile.PL has various code in it intended to be used for "bootstrap" purposes that we don't need to activate, and we're presently letting it trigger and then nuking bundled code after the fact. But it appears EUMM has ( since at least version 6.6 ) supported setting an ENV var "BUILDING_AS_PACKAGE" which turns it back into assuming it can be installed like a normal module, and this also disables bundling mechanics. Its clear this feature exists *for* vendors like us, so we may as well use it. https://metacpan.org/source/BINGOS/ExtUtils-MakeMaker-7.05_12/Makefile.PL#L21 https://metacpan.org/source/BINGOS/ExtUtils-MakeMaker-7.05_12/Makefile.PL#L48 And here is the line of code that does the bundled stuff that we'd be avoiding: https://metacpan.org/source/BINGOS/ExtUtils-MakeMaker-7.05_12/Makefile.PL#L78 Benefits include: 1. Potentially warning us about incompatible module scenarios that have been triggered by our avoidance of bundling 2. Running EUMMs own code and test suites against our vendor-provided versions instead of the bundled ones, helping further be assured that it can be expected to work after installation. 3. Eliminating some of the redundant code wrapping in src_install
Also, as articled in bug #541544, "Delete it if its not ExtUtils" is not sufficient for detecting Bundled things. Because ExtUtils::MakeMaker bundles ExtUtils::Manifest!
27 Feb 2015; Andreas K. Huettel <dilfridge@gentoo.org> +ExtUtils-MakeMaker-7.40.0-r1.ebuild: Fix bug 539996 and bug 541544