[Improvement] Separate SemVerSource from CommitCountSource

[asbjornu]

Currently, the term VersionSource is used both for calculation of the SemVer and for commit counting. The VersionSource output variables all point to the the source used for commit counting, not the source used for the resulting SemVer.

Detailed Description

The conflation of commit count source and SemVer source was introduced in #478, with the reasoning given in #465 that to avoid resetting the commit count, the oldest tag is used as the source for commit counting. The oldest tag is however not used as a source for the resulting SemVer, leading to the codebase operating on two different version sources.

Context

The conflation of commit count source and SemVer source is confusing, as #1830 and #2394 is proof of. A way to remove this confusion is to split VersionSource into two different concepts: SemVerSource and CommitCountSource.

Possible Implementation

The entire codebase needs to be refactored and every mention of VersionSource needs to be renamed to either SemVerSource or CommitCountSource depending on its usage.

[wsy-cloud]

Wonder if this would solve #2795 ?

[HHobeck]

Hi @asbjornu.

I would like to understand the problem and viewed the initial bug #465 and the actual fix for that #478 and tried to create an integration test which reproduce this issue. Could you please take a look and tell me what the expectation is? And also what was generated before this bug fix.

    [Test]
    public void __Just_A_Test__()
    {
        var configuration = ConfigBuilder.New.Build();

        using var fixture = new EmptyRepositoryFixture();

        fixture.Repository.MakeATaggedCommit("1.2.0");

        fixture.BranchTo("develop");
        fixture.MakeACommit(); // one
        fixture.MakeACommit(); // two

        // ✅ succeeds as expected
        fixture.AssertFullSemver("1.3.0-alpha.2", configuration);

        fixture.Checkout("main");
        fixture.BranchTo("hotfix/1.2.1");
        fixture.MakeACommit(); // three
        fixture.MakeACommit(); // four
        fixture.ApplyTag("1.2.1-beta.1");

        // ✅ succeeds as expected
        fixture.AssertFullSemver("1.2.1-beta.1", configuration);

        fixture.Checkout("main");
        fixture.MergeNoFF("hotfix/1.2.1");
        fixture.ApplyTag("1.2.1");

        // ✅ succeeds as expected
        fixture.AssertFullSemver("1.2.1", configuration);

        fixture.Checkout("develop");

        // ✅ succeeds as expected
        fixture.AssertFullSemver("1.3.0-alpha.2", configuration);

        fixture.MergeNoFF("hotfix/1.2.1");

        // ❌ expected: 1.3.0-alpha.2 or 1.3.0-alpha.3 or 1.3.0-alpha.4 or 1.3.0-alpha.5 ???
        fixture.AssertFullSemver("1.3.0-alpha.???", configuration);
    }

[asbjornu]

What GitVersion did before #478 is impossible for me to answer. I have no idea. 🤷🏼 When it comes to the test you've written, I think 1.3.0-alpha.5 makes sense because develop has 2 direct commits, hotfix/1.2.1 has two direct commits and then develop has the merge-commit from hotfix/1.2.1 as well. That amounts to a total of 5 commits on develop, 3 more than what develop had before the merge and the version was successfully asserted to be 1.3.0-alpha.2.

[HHobeck]

And this is the reason why the bugfix has been implemented and you want to separate the SemVerSource from CommitCountSource? Means in this example the SemVerSource is 5 and the CommitCountSource is 3? Or what is the your expectation?