Always include end of commit range

[wagoodman]

Currently there is erroneous logic for including or excluding the end of a commit range for a release (the end should always be included). The following diff is what this PR logically achieves:

diff --git a/chronicle/release/releasers/github/summarizer.go b/chronicle/release/releasers/github/summarizer.go
index c3f2406..f0cfedf 100644
--- a/chronicle/release/releasers/github/summarizer.go
+++ b/chronicle/release/releasers/github/summarizer.go
@@ -105,7 +105,7 @@ func (s *Summarizer) Changes(sinceRef, untilRef string) ([]change.Change, error)
        var changes []change.Change
        var err error
 
-       var includeStart, includeEnd bool
+       var includeStart bool
 
        var sinceTag *git.Tag
        sinceHash := sinceRef
@@ -135,13 +135,11 @@ func (s *Summarizer) Changes(sinceRef, untilRef string) ([]change.Change, error)
                if err != nil {
                        return nil, err
                }
-               includeEnd = false
        } else {
                untilHash, err = s.git.HeadTagOrCommit()
                if err != nil {
                        return nil, err
                }
-               includeEnd = true
        }
 
        var includeCommits []string
@@ -150,7 +148,7 @@ func (s *Summarizer) Changes(sinceRef, untilRef string) ([]change.Change, error)
                        SinceRef:     sinceHash,
                        UntilRef:     untilHash,
                        IncludeStart: includeStart,
-                       IncludeEnd:   includeEnd,
+                       IncludeEnd:   true,
                })
                if err != nil {
                        return nil, fmt.Errorf("unable to fetch commit range: %v", err)

However to properly put this change under test additional functional decomposition was necessary. Since this range selection was refactored to be more accurate it additionally considers the case where no tags were found, in which case the first commit in the repo is used.

Partially addresses #96

[willmurphyscode]

Nit: nope.com is a real domain name that resolves. (I don't know what's there; I just checked the NS lookup). Should we instead use [email protected] or something?

[wagoodman]

I can make that update, but that affects more than just these new fixtures. Can we follow up with another PR that makes that adjustment universally? (I don't see a direct impact here so will leave these values as is for now)

[willmurphyscode]

Definitely ok with splitting it up. I don't think this is super urgent, since we're never going to actually email these folks, but I remember hearing that the owners of test.com or similar are very tired of getting bot spam from everyone's QA accounts, and example.com is the canonical valid domain name that no one uses.

[willmurphyscode]

This reassignment and naming confused me a little bit. It looks like you make a real gitter, and a mock gitter, but the mock gitter has a pointer to the real gitter, and you can assign either the mock or real gitter to a variable called gitter. Maybe the comment here could be rephrased, or one of the gitter vars could get a different name?

[wagoodman]

I'll update the comment, using the same name is intentional: any usage of the unmocked gitter would be invalid from that point forward.

[willmurphyscode]

I want to make sure I understand here. Inclusive means "the changelog should include this point"? And then below, when we generate a changelog that includes the first commit in the repo, we need to emit an inclusive change point (because the first commit is part of the first release), but subsequent releases, where we're asking for something like "commits since the tag v0.5.1", we should make an exclusive range. Is that right?

[wagoodman]

that's 100% correct 👍

[willmurphyscode]

I don't understand what this config does.

[wagoodman]

looks like this option never was added to the documentation, I can update that in a separate PR (here's the PR that put this feature in #40 ) . In short, when enabled the commits that comprise the release can affect which PRs get included into the release notes.

[kzantow]

LGTM but I'll leave it to @willmurphyscode to approve, since this is being looked at by both of us