WIP: refactor: use git-fs-repo instead of git cli

[marionebl]

• introduces git-fs-repo
• removes reliance on git cli
• replaces regex parsing of git log with iteration over [^/refs/tags/*]
• retains sorting of results of previous versions

---

• depends on guard transform_plain_ref against empty buffers chrisdickinson/git-load-refs#3 for repositories with empty refs
• investigate bugs related to git gc
• test case for shallow clones
• test case for repositories with dangling refs
• investigate bugs related to commits without author information in git-walk-refs

I had some issues with the current RegExp technique skipping some git
tags in one of my projects.

I am still investigating this and a second, short-term mergeable PR will land tomorrow.

While dabbling around with js-only git implementations the current PR came into being. This is
meant as basis for discussion about a move to a js-only conventional-changelog.

[coveralls]

Coverage decreased (-3.9%) to 96.124% when pulling efa1422 on marionebl:master into 5390b10 on stevemao:master.

[coveralls]

Coverage decreased (-3.9%) to 96.124% when pulling dd79ad3 on marionebl:master into 5390b10 on stevemao:master.

[coveralls]

Coverage decreased (-4.5%) to 95.455% when pulling 87bac03 on marionebl:master into 5390b10 on stevemao:master.

[stevemao]

Thanks for the PR!
+1 for js-only git implementations.
Can you annotate the code? Since it's not simple logic anymore and it will increase the barrier for people to contribute.

I had some issues with the current RegExp technique skipping some git
tags in one of my projects.

What exactly is the issue?

[coveralls]

Coverage decreased (-4.5%) to 95.455% when pulling bacf735 on marionebl:master into 5390b10 on stevemao:master.

[marionebl]

@stevemao added JSDoc and inline comments for everything. I'll open a distinct PR for the RegExp issues, stay tuned.

The failing coverage threshold is caused by uncovered callback error handling branches. Any ideas?

[coveralls]

Coverage decreased (-4.3%) to 95.652% when pulling 813cbe4 on marionebl:master into 5390b10 on stevemao:master.

[coveralls]

Coverage decreased (-4.3%) to 95.652% when pulling 02486a8 on marionebl:master into 5390b10 on stevemao:master.

[marionebl]

anything I need to do before you can take another look on this?

[coveralls]

Coverage remained the same at 100.0% when pulling 0f29b9a on marionebl:master into 5390b10 on stevemao:master.

[coveralls]

Coverage remained the same at 100.0% when pulling f319032 on marionebl:master into 5390b10 on stevemao:master.

[marionebl]

Holding back for now, as I am investigating some stability issues related to git-walk-refs

[marionebl]

Dropping this as I don't have time to go forward on making the js-only implementation reliable

[stevemao]

No need to close a issue if you don't have time. Anyone who's interested can pick this up

[coveralls]

Coverage remained the same at 100.0% when pulling f319032 on marionebl:master into 5390b10 on stevemao:master.

[JaKXz]

I might be able to pick this up since I'm interested in it. I'll give it a whirl this weekend and try to have an update.

[JaKXz]

@stevemao unfortunately I didn't have much time - but I did get the branch rebased just fine and all the tests pass. Want me to make the PR? I didn't pick up enough context in the few minutes I looked at it so I can't say for sure that it's ready to merge.

[stevemao]

Sure, thanks for the contribution :)

[bcoe]

@stevemao @JaKXz @marionebl should we consider opening this issue up on the new monorepo?

[hutson]

We're closing this pull request out as this repository has been merged into our conventional-changelog mono-repository.

If you would like to see this change incorporated into conventional-changelog please consider re-opening the pull request against the conventional-changelog mono-repository.