Impact is a crash detection and recording library for Apple platforms. It is not a full crash reporting system. But, it could be the core of one. Its design goals are:
Current feature set:
Impact uses a text-based log format. While the format itself is stable, the contents are still unstable. It is designed to be simple, while still making both debugging and parsing possible in the face of crash-time failure.
In-process crash reporting is just terrible. The mechanisms available for crash event detection, UNIX signals and Mach exceptions, are complex, buggy, and are not capable of capturing all kinds of failures. On top of that, the enviroment in which a crash reporter needs to run is extraordinarily hostile. It's just messy business.
Apple has had all of the device-side pieces in place to produce a world-class crash reporting system for a long time. While they have the reporting side down, the developer experience (analysis, presentation, investigative tools) leaves a lot to be desired. This has kept 3rd-party reporting services essential for the vast majority of app developers. Apple's system also doesn't work for macOS apps outside the App Store, which is disappointing.
My sincere hope is that Apple addresses these limitations so we can all stop this foolishness once and for all.
But, for now, in-process reporting is a necessary component for most developers. There are many trade-offs and design decisions that dramatically affect the qualities of a reporting system. Understanding those trade-offs, and being explicit about choices that affect them is one of the goals of this project.
Also, crash reporting is just a fun and fascinating problem. It tends to be very commonly used and very poorly understood. I think the Apple development community could benefit a lot from a more in-depth understanding of the area.
You must keep in mind that Impact only captures information about crash events. It does not have any facilities for transmitting those events back to you or translating them into human-readable versions. A full reporting system requires a little more work. Here are a few options:
I worked at Crashlytics for many years. During my time there, I briefly worked with PLCrashReporter before starting from scratch and building a completely custom system. I spent a considerable amount of time there analyzing and understanding the failure modes of in-process crash reporting. That work shaped most of the design of the Crashlytics SDK, though things might have changed since I left.
Impact does share many of those design philosophies. It's hard to unsee solutions, sometimes. But, it's primarily because I believe those core design concepts are the best approach.
It would be wonderful to see contributions. If you'd like to work on something, the safest bet is to open an issue or PR first. That way, we can discuss the changes before you spend too much time working.
There is absolutely no experience/knowledge requirement. Interest is all you need, I am happy to help.
We'd love to hear from you! Get in touch via twitter, an issue, or a pull request.
Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.
|Last commit: 1 week ago|