Win fix

[wchristian]

This fixes various tests that unnecessarily fail on windows.

[petdance]

Neither @hoelzro nor I understand why the SKIP blocks are there. Why are those tests dying on Windows? I'd rather fix those problems than just SKIP past them.

[wchristian]

Where are you discussing this? I tried poking @hoelzro on Twitter and he didn't react, and neither of you are on IRC.

As for your question, those tests die because of this particularly unfortunate instruction: https://github.com/petdance/ack2/commit/1e09ac8f9150ecfca836d54ab25f55084f1ebe5c#L0R165

I'm honestly confused neither of you are aware of this, given the copious amounts of cpantesters FAILs this has generated: http://matrix.cpantesters.org/?dist=ack%202.08;os=mswin32;reports=1

As for a fix, i honestly didn't try too hard to understand what that code is attempting to do, but maybe this might help circumvent some shell issues? https://metacpan.org/module/ExtUtils::Scriptlet

[petdance]

Twitter is not a good tool for conversations. Both of our email addresses are easily findable. Mine is in my Twitter profile if nothing else. And no, I'm not on IRC. The best place to discuss is probably the ack-dev mailing list on groups.google.com.

Yes, we're aware of the failures, but it is also singularly difficult to test anything on Win32 as I have no Windows anything at my disposal.

[petdance]

Also, I have to confess that it's very difficult for me to slog through CPAN testers failures, because so many of them are false.

[wchristian]

That doesn't answer the question posed. Either way, does my answer tell you enough to convince you that this is good to merge or do you need more info?

[petdance]

Which question are you referring to?

[wchristian]

Oh, also, i'll happily give you VNC/rdp access to a win7 machine to poke.

[wchristian]

This question:

Neither @hoelzro nor I understand why the SKIP blocks are there.

Where are you discussing this?

[petdance]

Re: "Where are you discussing this?" We've just been discussing in AIM.

And a VNC/rdp would be a huge help.

[hoelzro]

@wchristian Sorry for not getting back to you sooner; I didn't see your Twitter message until you pointed it out just now. Which IRC channel(s) are you checking? I'm usually on #perl on Freenode.

Anyway...the reason this PR is under debate because of all of the SKIPs it adds. I'd personally rather have it fail outright.

Andy and I are aware of the Windows problems, but speaking for myself, I don't have a Windows environment with which to work.

We should definitely discuss this more on ack-dev.

[wchristian]

I've made a post on the ack-dev list which will be visible once it's approved.

[shadowcat-mst]

It would be better to have the feature die with an explicit error when invoked on Win32 - and maybe test for that - than to cut off the entire ack+Win32 userbase just because you haven't had the time/resources to implement one feature on windows yet.

Up until this die() was added ack would install on windows, and the basic functionality worked fine.

So failing outright is achieving nothing except for punishing existing users who, other than this, have a perfectly working program that they'd like to still be able to install on new systems.

Perhaps a good middle ground would be to mark the tests as TODO on win32, so they still run (and produce failure information), but don't stop installation? That would leave them in a state where they could be worked on without punishing your existing user base.

[wchristian]

Changed the pull request to convert most of the SKIPs to TODOs, except for passthru, which would go into an endless loop.

[hoelzro]

@wchristian Thanks for the change! I'm going to merge this, since the die() thing is a pretty serious problem, and I'm a lot more keen on TODOs for this.