Alternative Include-default tests

[kentfredric]

This branch incorporates the patches to Manifest.pm by @mohawk2 in pull #17.

However, it takes a different approach at testing, trying to put our best foot forwards in terms of not making the existing tests worse.

This sequence is of course, presently advisory, and commits can be squashed as needed.

Ideally I would make the file IO inside the test-case neater by having a tinyer version of Path::Tiny so you can more clearly see where the tests are happening and where the cruft is happening.

Additionally, this code would be substantially neater if dagolden/Test-TempDir-Tiny#7 was done.

My hope is this test will be 100% clear in exactly what it is testing, and why it is testing it.

There is by comparison nothing clear about: https://github.com/mohawk2/ExtUtils-Manifest/commit/4ab7d94a330848fb20d02bcbd5a5928d57ed3a9d#diff-58b9e890377f53ed47933c81a658843fR439

[mohawk2]

I like what you've done here. How close are we to being ready to merge / release?

[kentfredric]

I've rebased this replacing the T:TD:T 0.004 with the pre-release of 0.005 ( w/ local merging ).

Effective dependency changes in the commit message.

the tests itself has been slightly restructured so the spew/slurp stages don't complicate understanding "What the test does"

[kentfredric]

Its also possibly worth noticing that this patch by @mohawk2 does NOT change the behavior of "!include" ( Which it probably should ), but only changes the behavior of "!include_default".

I'll gladly attempt to fix that when this patch gets approved as changes relative to this one ( But It will necessitate adding a module with slurp/spew in it to share it nicely.

[kentfredric]

Fun fact: Looks like I can stick File::Slurper in tlib/ without creating any more dependencies than Test::TempDir::Tiny already does, and it greatly simplifies code without needing me to re-invent and re-write how slurping works:

https://github.com/kentfredric/ExtUtils-Manifest/blob/fileslurper/t/maniskip.include-default.t

[kentfredric]

NB: Githubs commit ordering is completely bollocks now.

29d7f44 Add Test::TempDir::Tiny 0.005-pre-release verbatim in t/tlib.
61d5833 Simplify unnecessary POD that wont get installed
f4e98f8 Add tlib/ByteSlurper.pm for reading/reading files as bytestrings
91b879e Add mailmap to map kentnl's CPAN email address
02d3e01 Doc, implement #!include_default as memory-only, not changing file.
d55d35a Add isolated tests for Mohawks include-default patch

This order is much more logical.

I've rebased/squashed things several times now:

Recent changes primarily include:

• The factoring out of the spewing/slurping logic into its own module for re-use later
• Nuking the unwanted POD from the stolen T:TD:T for a slimmer install ( and avoiding invalid POD ).

The "Nuke POD" change was done in an independent commit to make future updates to this module easier where necessary.

[mohawk2]

I think #!include is probably intended to incorporate author-made contents for use at install-time. #!include_default is special because one can safely assume that at install-time, "default" will be available (and can change over time which is the motivation for the change), whereas separate author-made stuff won't be.

[haarg]

MANIFEST.SKIP doesn't matter at install time. It's only used to build the MANIFEST.

[mohawk2]

Ah, that's right! #!include is also author-side only. Should I change that too?

[haarg]

@mohawk2 Yes, definitely.

[kentfredric]

NB: I saw the IRC comment on this issue, but the box I run my irc stuff has now succumbed to hardware issues and no longer POSTs, so my IRC contact might be a little bit before its back.