Add Raven.showReportDialog (experimental)

[benvinegar]

No description provided.

[dcramer]

dsn needs to be configurable via options

[mattrobenolt]

We probably shouldn't throw errors like this since they'd then get caught and in turn reported to Sentry. Thoughts?

Or we have a rule to ignore RavenConfigError's. This wasn't an issue before because RavenConfigError was only ever thrown before raven was even hooked up. Now we'll be throwing errors that will be caught and logged.

Not sure if bad or not, just pointing it out.

[benvinegar]

Wouldn't you want them reported to Sentry? It means that you misconfigured Raven (for dialog purposes) and you might never otherwise find out.

[benvinegar]

Note that if you have Raven configured to accept an error, you will have set a DSN, and will thus never hit Missing DSN.

[mattrobenolt]

I could see validity in the Missing DSN case, since that's before Raven is configured, but for missing event id... I'm not sure. I feel like we could just return a true/false instead for success/failure?

[benvinegar]

I feel it should be reported.

[mattrobenolt]

[benvinegar]

I think it's really hard to get into an infinite loop situation here, because when calling showReportDialog:

1. If you don't have a DSN, then Raven wasn't installed and the global error handler isn't present – the exception is thrown but nothing catches it and control flow ends.

2. If you don't have an eventId (or there was no lastEventId), then Raven will throw an exception, and this will be caught by the global error handler. Now if showReportDialog is called a second time, now there is a lastEventId and the function will no longer error.

Also, if you have an onerror handler, and inside that handler you throw another error (e.g. RavenConfigError), onerror doesn't get called again in an infinite loop – it just stops there:

http://jsbin.com/qoqoviduru/1/edit?html,js

The user would have to use a setTimeout call to leave the onerror stack to reach this infinite loop scenario:

http://jsbin.com/moxulabigi/1/edit?html,js,output (warning, this will freeze your browser)

But I mean, if they did this, any kind of bug in their code would trigger the infinite/setTimeout scenario – not strictly this RavenConfigError. So altogether, I think this scenario is really unlikely, and would be their fault anyways.

I'm going to move ahead with merging the branch. It's flagged experimental anyways.

[stiller-leser]

Hey,

first off: Thanks for the implementation. Looks like this is going to help us a lot. I know this is a beta-feature, but I thought I'd give it a shot anyways: Is there a way to localize the error-embed? It did not look like setting a locale via Raven.setUserContext would do the trick.