Version bumping not working

[sebbacon]

Despite #179 appearing to work when testing, releases are not being generated as expected.

Here is a run that does not consider this a release. This appears to be because the merge commit does not include the relevant conventional commit type - i.e. it only seems to be considering the merge commit.

However, the documentation makes it clear that:

The action will parse the new commits since the last tag using the semantic-release conventions

However, the last tag, v0.1.29, is at commit 49aadc2, since when there have been two commits starting fix:.

The command that the build linked to above runs is:

git log v0.1.29..HEAD --pretty=format:'%s%n%b|commit-hash:%h==============================================' --abbrev-commit

And its output on master for me is:

$ git log v0.1.29..HEAD --pretty=format:'%s%n%b|commit-hash:%h==============================================' --abbrev-commit
Merge pull request #231 from opensafely/test-release-making
Tweak docs and in doing so test release-making|commit-hash:99bb748==============================================
fix: remove broken reference to roadmap in docs
This is not really a `fix`, but I want to test triggering a new release.
|commit-hash:c9bb852==============================================
Merge pull request #230 from opensafely/semantic-autoversioning
fix: use semantic versioning automation|commit-hash:718d063==============================================
Merge branch 'master' into semantic-autoversioning
|commit-hash:74ff162==============================================
docs: bring description of release process up to date
|commit-hash:08b8304==============================================
fix: use semantic versioning automation
This is not really a fix, but we do want to trigger a new release.

Closes #179
|commit-hash:6581bf0==============================================

Given the run log ends

Tweak docs and in doing so test release-making
The commit should not trigger a release
Analysis of 1 commits complete: no release

It seems odd that it should consider an analysis of 1 commit to be complete.

[sebbacon]

There is an issue reporting the same problem for the action here.

And here is someone working round a similar problem by rebasing rather than merging

[sebbacon]

I presume the bug is in this underlying library. I suspect this bug because our merge commits contain slashes.

[sebbacon]

The problem is that the current version of the checkout action checks the code out without any history (depth=1)

[inglesp]

The problem is that the current version of the checkout action checks the code out without any history (depth=1)

Well done for getting to the bottom of that!