Git存储库结构和工作流程

Question

How best to structure our Git and workflow? I am pretty new to managing Git in a multi user environment and come from a Perforce background so I was wondering if anybody can help

33k+ database objects spead across approx 1000 distinct functional areas.

300 users working at the same time on 17 projects.

Each project has its own development cycle using a mix of waterfall and scrum\\agile.

Typicaly a project touches 6 to 10 functional areas though some of these areas are updated almost all the time

Want to move to a fortnighly release into production and become more agile

We want to track versions at a low level so as to help the RM's track changes and to identify and communicate changes in dataflow and areas of responsibility

Projects work with local database systems but for regression and performance testing they use 2 shared environments that must be treated the same as production

We want to introduce continuous build and continuous regression testing

I propose to introduce a seperate repository for each functional area

I propose to use Gitflow Workflow

I propose to modify the Gitflow scripts to make managing features across multiple repositories part of its behaviour

I have read a lot about submodules and subtrees but dont really think either of these are suitable as we do everything in multiple branches.

Any sugestions or pointers would be much appreciated

Answer 1

We use this workflow on our projects, with some difference:

• I prefer rebasing rather than merging when putting a branch in develop , because with squash we can have a single commit per issue, giving a more clear and linear history.
• We use when needed some pre-push hooks in order to perform unit test before pushing our work into remote, in order to avoid that other developer will have problems with our code.

Also in your case I would prefer submodule over subtree, because subtree set in a repository all source code of sub-repository, while submodule take it apart.

I'd create all repositories that you need. Developer should develop on a repository at time, and there must be some supervisor that integrates single repository in bigger ones, managing and handling issues.