Add centralized error reporting via Rage::Errors

[Digvijay-x1]

Allow users to register error reporters that receive all framework-caught exceptions, so they can forward them to services like Sentry or Bugsnag without manually rescuing in every component.

  Rage.configure do
    config.errors << SentryReporter.new
  end

Reporters can accept just the exception or an optional context hash. Signature detection uses Rage::Internal.build_arguments at registration time to keep the reporting path fast. Each reporter is called in its own rescue block so a broken reporter doesn't affect the rest. Duplicate prevention uses an instance variable on the exception (similar to Rails). Manually-created exceptions get a real backtrace via re-raise rather than set_backtrace(caller).
Hooked into all existing framework rescue sites: controllers, fiber wrapper, Cable connections/channels, Redis adapter, deferred tasks, event subscribers, SSE streams, and telemetry tracers.

Fixes #232

[rsamoilov]

Hey @Digvijay-x1 ,

I usually don't review in-progress PRs, so I haven't looked at this one yet. Anything I can help with? Do you want me to give it a look?

[Digvijay-x1]

Hey @Digvijay-x1 ,

I usually don't review in-progress PRs, so I haven't looked at this one yet. Anything I can help with? Do you want me to give it a look?

Hey @rsamoilov,

Thanks for offering to help :)

Before requesting a review, I wanted to test whether all cases behave as expected. This took a bit longer since I had some college assignments last week.

From my testing so far, the core error reporting API appears to be working well. I set up a small Rage app with Sentry integration and verified that reporter registration, error forwarding, context propagation, duplicate prevention, and backtrace generation all behave correctly.

I’ve tested 5 out of the 9 call sites so far — including unhandled controller exceptions (covering both application.rb and fiber_wrapper.rb), deferred task failures, and direct Rage::Errors.report calls.

The remaining areas (events/subscriber, SSE, Cable, Redis adapter, telemetry) all route through the same Rage::Errors.report method, which is already working as expected.

I think the PR is in a good state for review now. Would appreciate your feedback whenever you get a chance!

[rsamoilov]

Hey @Digvijay-x1 , the code looks good. I'll test the implementation more thoroughly and get back to you early next week.

[rsamoilov]

Hi @Digvijay-x1 ,

This is an amazing implementation. I love how minimal it is - this is exactly what I had in mind.

My main comment is that you're mixing two different concerns (configuration and reporting) inside Rage::Errors.

report should be part of the module - that's OK. But configuration (<<) belongs to Rage::Configuration - you should have a separate namespace there that allows to add and remove error reporters. For example:

Rage.configure do
  config.error_handlers << MyErrorHandler.new
end

Then, once a reporter is added, the configuration code should call a @private method on Rage::Errors to actually register the reporter.

[rsamoilov]

Suggested change

	Rage.logger.error("Error reporter #{@reporters[i].class} failed: #{e.class} (#{e.message})")
	Rage.logger.error("Error reporter #{@reporters[i].class} failed while reporting #{exception.class}: #{e.class} (#{e.message})")

[rsamoilov]

Let's also add the reporter hook into Rage::PubSub::Adapters::Redis (that's a new module that will replace Rage::Cable::Adapters::Redis) and add a shortcut to access Rage::Errors via Rage.errors to the main module.

[rsamoilov]

Looks good! Can't wait to include this into the next release!