add error_stack on reject also reject error not string

[dreadjr]

Wanted to start a discussion about 2 things,

1. adding error_stack
2. reject new Error() vs string

[googlebot]

Thanks for your pull request. It looks like this may be your first contribution to a Google open source project, in which case you'll need to sign a Contributor License Agreement (CLA).

:memo: Please visit https://cla.developers.google.com/ to sign.

Once you've signed, please reply here (e.g. I signed it!) and we'll verify. Thanks.

---

• If you've already signed a CLA, it's possible we don't have your GitHub username or you're using a different email address. Check your existing CLA data and verify that your email is set on your git commits.
• If you signed the CLA as a corporation, please let us know the company's name.

[dreadjr]

I signed it!

[googlebot]

CLAs look good, thanks!

[cbraynor]

I'll take a look at this today - initially I used strings because they are easiest to represent in Firebase, and perhaps we don't want to coerce every rejection with a string into an Error object, but a stack trace would be a useful addition

[dreadjr]

Ok cool, yeah i wasn't sure about the error stack as far as privacy/security is concerned either. It might need to go in an "errors" bucket and just reference it. I was thinking could just do a JSON.stringify as well.

[cbraynor]

In using the queue, I've tended to lock down the /tasks location to write only (and create only) for clients that push items onto the queue, and only allowed the trusted server processes read access. This means that for me, adding the stack trace to the task is a non-issue.

I'm curious whether that's a pattern other people are using, or whether read access is indeed required in some use-cases? If read access is required, then you're right that adding stack traces could be a security concern

[dreadjr]

Yeah true. I was thinking too of converting stack trace into Jain so it could be saved to firebase more easily.

[cbraynor]

I included these changes in #24 with a few minor adjustments. The major changes were

1. We don't specifically coerce all errors to Error objects - if they're strings then they behave the same as they do now (without a stack trace)
2. There's now a suppressStack option for the queue that prevents stack traces from ending up in the Firebase

Take a look and let me know if there's something I missed. Let's move the conversation there