GitVersionTask doesn't fail the build when tip of head isn't a merge commit

[andreasohlund]

When this happens no version is applied and it might be hard to debug since its not obvious that it failed

[distantcam]

For me, an error that stops the build should only occur for something that the user can control.

Yes, the user can fix it, but it requires creating a new commit.

However, if the user is building an older commit, tip of head cannot be modified so again, the user is stuck and unable to build without digging out GV.

Ultimately I'm -1 to this.

[pvandervelde]

Technically the same will happen if you have a completely new repository with no commits on master. Again you can fix it but why should the user have to make a merge commit on master for it to work?

[distantcam]

@pvandervelde This error only occurs when there's a develop branch and GitVersion is in GitFlow mode. A new repository would be GitHubFlow mode.

[pvandervelde]

Ah right. So how does one start a new repository in GitFlow mode correctly then? I've worked with GitHubFlow fine but I just can't seem to get a new repository to work with GitFlow mode. Is there anything specific I need to do in order to make it work?

[robdmoore]

There needs to be a branch called develop from memory.

[pvandervelde]

Does that branch have to have anything 'in it'? As in does it have to be different from master. So can I do:

• Create repository
• Add files to master branch
• Branch develop from master

Would that work or does develop have to have at least one commit on it? I suspect it does not because it's always worked on my local machine (where all the branches are present in the local workspace). It just doesn't work on Jenkins where only the required data is fetched (see #179)

[JakeGinnivan]

This is not a problem in 3.0.0 branch