Crash ping data isn't our data set, but Socorro-style signature generation is. There's a need for generating Socorro-style signatures for crash ping data. That requires two steps:

1. symbolicating the crash ping stack (Tecken)
2. running the symbolicated stack through signature generation (Socorro)

Last summer, there was work done on this which produced fx-crash-sig (https://github.com/Ben-Wu/fx-crash-sig) which apparently is in use. It uses a very old version of siggen which was our experimental extracted Socorro-style signature generation library which also hasn't been maintained. The problem is that signatures generated with that don't benefit from all the updates and changes we've made in the last year nor any future changes.

This bug covers figuring out what to do in the short term so that the WebRender group can generate signatures for crash ping data.

Making this a P1. It's weird that it's a Socorro bug, but I expect we'll probably re-home it and generate a few bugs out of it.

Will: What's the status with Ben Wu's work or any other work for generating Socorro-style signatures for crash ping data?

Ben's project is on hold for now. Data science is doing some investigations into our stability program in general as it pertains to release management (they will probably be contacting you soon if they haven't already) and I'd like to get their opinions/results before committing to doing any more large-scale work.

That said, I'm happy to help with improving/maintaining these libraries/tools if people have short term goals like a stability dashboard showing the top N crashes in the field (I heard from Jeff that Kats had already prototyped something like this). That sounds super useful and hopefully not too much work to build and maintain.

That makes sense.

I'll update siggen and make sure I continue to update it going forward whenever there are signature generation changes in Socorro. I'll write up some issues against fx-crash-sig for things that need to get fixed. I think that'll address maintenance issues. Then whoever uses that fx-crash-sig will be responsible for making sure they've got the latest fx-crash-sig and siggen.

We'll see how that goes.

I updated siggen and pushed out 1.0.0.

I added a couple of issues to fx-crash-sig and threw together a pull request. I don't have admin access to that repo and I don't know who does. Ben, I presume does. I'll work with him to land that and get a new version out.

(In reply to Will Kahn-Greene [:willkg] ET needinfo? me from comment #4)

I updated siggen and pushed out 1.0.0.

I added a couple of issues to fx-crash-sig and threw together a pull request. I don't have admin access to that repo and I don't know who does. Ben, I presume does. I'll work with him to land that and get a new version out.

We should probably get Ben to transfer that org to Mozilla and set you, me, and him as admins.

That's a good idea. Plus we'll need access on PyPI to do releases. Fun fun fun!

I wrote up an issue covering those two things: https://github.com/Ben-Wu/fx-crash-sig/issues/6

(In reply to Will Kahn-Greene [:willkg] ET needinfo? me from comment #6)

That's a good idea. Plus we'll need access on PyPI to do releases. Fun fun fun!

Thanks! I believe I already have admin rights to the pypi project, so that angle is covered.

We transferred fx-crash-sig to the mozilla GitHub org and some of the PRs landed. I think we're either good for a release or something.

So, this is a short-term fix for figuring out signature generation for crash ping data. We probably want a long term fix, but that's outside the scope of this bug and can be handled in another bug in another castle.

Marking this one as FIXED.