653313 - Several Validator stack traces

Status: RESOLVED FIXED

(addons.mozilla.org Graveyard :: Admin/Editor Tools, defect)

Description

[Kris Maglione [:kmag]]

User-Agent:       Mozilla/5.0 (X11; Linux x86_64; rv:6.0a1) Gecko/20110419 Firefox/6.0a1
Build Identifier: 

I'm attaching three sets of stack traces from three different add-ons. The review pages are linked (I can attach the XPIs if that helps). I'm seeing these in the HEAD of Matt's repo. Kumar asked that I file a bug for the first two, the others came up later (I've seen the stack overflow a few times in the past, too).

Reproducible: Always

Comment 1

[Kris Maglione [:kmag]]

Attachment: Stack traces

Comment 2

[Matt Basta [:basta]]

To the best of my knowledge, I've fixed all of these in the most recent version(s) of the validator. They were mostly fixed in the big push after the first round of changes to fix the functional tests.

Could you attach an XPI for the very last trace? I don't think I've seen that one before.

Comment 3

[Kris Maglione [:kmag]]

Attachment: Troublesome XPI

Comment 4

[Matt Basta [:basta]]

Awesome. That last bug is definitely new. I patched it and included it in the pull for the latest JS stuff. Could you also attach the first XPI as well?

I've already fixed the int/startswith bug, the recursion bug, the 'math' bug, and now the 'NoneType' bug. I figure that I'd better be safe than sorry with that bug at the beginning.

Comment 5

[Kumar McMillan [:kumar]]

Thanks Kris. It looks like the first two will be fixed soon with a merge but could you attach those XPIs so we can create regression tests?

Comment 6

[Kris Maglione [:kmag]]

Attachment: Troublesome XPIs

Other two failing XPIs.

Comment 7

[Matt Basta [:basta]]

These look like they're all fixed up in the most recent commits. Kumar, do you want me to keep this open so you can create regression tests?

Comment 8

[Kumar McMillan [:kumar]]

yeah, good idea

Comment 9

[Kumar McMillan [:kumar]]

Fixing this because it works! We are sort of backlogged on regression/functional tests but that will be remedied soon.

Comment 10

[Matt Basta [:basta]]

I'm running Diggy fairly regularly now, too, so having general regression tests for outright tracebacks is a lot less important.

Comment 11

[Kumar McMillan [:kumar]]

(In reply to comment #10)
> I'm running Diggy fairly regularly now, too, so having general regression
> tests for outright tracebacks is a lot less important.

The regression tests are to prevent breakage when the code gets refactored in ways that the unit tests don't catch.