Stuffz

[ribasushi]

This does not address any of the concerns raised in #1, I am going to think about this later today.

All the changs are just a proposal, I am happy to rework things myself and/or have you rip out pieces.

[ribasushi]

Force pushed a new version. The following parts of our conversation are adressed:

• I removed the "version unset" and replaced it with a subtle 'n/a'. After you mentioned that some modules do set an "undef" as string, I figured this is a sane compromise. If you still want undef - we can do that ;)
• subsumes bug was fixed: https://github.com/ribasushi/Test-DiagINC/commit/6bb51e5623179e84ecb8dcc31e448652e3a00c0b#diff-b56f485fc595edf35a1f800cc587e717R57 I am now really paranoid I am still missing something, please review :)
• Getting Cwd from a subprocess. $^X -MCwd -e print(Cwd::getcwd) doesn't work on plain linux shell, so I opted for a much more stable solution (see https://github.com/ribasushi/Test-DiagINC/commit/15bd668529a51f4017070e6ba57e877c7c5523ec#diff-b56f485fc595edf35a1f800cc587e717R22 ). There may still be issues with quoting of space-containing paths, will need to investigate.
• The logic of when to load B is rewritten, but the 2-pass approach was left in. It feels wrong to load B / install an extra END always if we know we already failed anyway. I still can amend it if you insist there be just one "at the very end" pass. Note that the diag report will be the same either way - we will only report stuff from the %INC snapshot, not anything we load afterwards.
• The docs for "load this early" were ammended here https://github.com/ribasushi/Test-DiagINC/commit/cf216bcaaaf74c30a45ccf6d4330742f4fe46d5e#diff-b56f485fc595edf35a1f800cc587e717L94 and here https://github.com/ribasushi/Test-DiagINC/commit/cf216bcaaaf74c30a45ccf6d4330742f4fe46d5e#diff-b56f485fc595edf35a1f800cc587e717L145 . Feel free to ask for MOAR tweaks ;)
• The load failed for the time being is just a plain string, no extra logic. I agree with the EUMM walk, and I will do it, but it seemed like a good candidate for a separate PR, once we get the current stuff smoked better. As of now all I do is: https://github.com/ribasushi/Test-DiagINC/commit/528faa5d577098354e99a3c38cb700dbb5298e27#diff-b56f485fc595edf35a1f800cc587e717R92

Cheers

[ribasushi]

Re Getting Cwd from a subprocess - @haarg suggested a variation of the oneliner-getcwd invoke on #perl-qa, look into it before deciding the final implementation.

[ribasushi]

Folks dislike the -c hack, abandoning this PR in lieu of #4