What does homu do?
It works by reordering the common way that continuous integration is done. Normally, it goes in this order:
- A developer’s “working branch” is uploaded to the shared repository.
- A reviewer checks that it’s good, and, when satisfied that it is, merges it into the “main branch” and uploads it.
- A server cluster, running all of the OS’s that the software product needs to run on, downloads the main branch and tests it.
- If the cluster says it passed, then it’s probably suitable
- If the cluster says it failed, the developers are expected to fix it. Now.
The problem is that the software product only gets tested after it’s already been uploaded to the main branch, and since other developers are expected to use that branch as the basis for their own work, they’ve started basing their work on a bad branch.
Homu does it this way:
- A developer’s “working branch” is still uploaded the same way.
- A reviewer checks that it’s good, and, when satisfied that it is, sends a message to homu.
- Homu merges it together with the main branch, but uploads it to a separate branch (typically called “auto”).
- A server cluster, running all of the OS’s, downloads the auto branch and tests it.
- If the cluster says it passed, homu responds by copying “auto” into “main” (verbatim, since auto is always main-plus-a-change).
- If the cluster says it failed, homu reports the error but does nothing else.
This helps when there are lots of developers, because the main branch is always a copy of auto that passed. If reviewers approve changes more quickly than homu can build them, it will work through them one at a time in the order it receives them.
What’s going on with it?
It doesn’t exist any more. @barosl hasn’t updated it for years, the homu.io domain was allowed to expire, and nobody else has the source code for the frontend. You can still run the OSS version of homu if you want, like Servo and Rust are doing.
The Rust fork is considered the “canonical” version that you should use, since they’ve been adding new features and fixing bugs in homu. If you want to run your own instance, Rust’s fork also has a Dockerfile, which they’re running on top of Amazon ECS and terraform.
So what’s bors?
It aims to be faster and more user-friendly while sticking to mostly the same idea. You can register for the publicly-hosted version run by @notriddle. If you want to set up your own instance, but are having trouble with the instructions, please ask about it on the forum, where the rules are “it’s better to just not answer the question than to be rude, even if it’s a question that’s been asked before.”