[test] compare Tecken symbolication API with Eliot symbolication API
Categories
(Eliot :: General, task, P2)
Tracking
(Not tracked)
People
(Reporter: willkg, Unassigned)
References
Details
Once we get Eliot into stage, we want to run a symbolication jobs through the existing Tecken symbolication API and the new Eliot symbolication API.
To build test stacks, we're going to convert existing crash reports. That gives us two sources of truth (what did stackwalker say? what did tecken say?) to compare with.
We'll perform requests against the two systems, capture the output, and compare. We should investigate differences.
One expected difference is that Eliot's v5 symbolication API has an additional "line" in each frame that isn't in Tecken's v5 symbolication API. We can ignore that difference.
|
Reporter | |
Comment 1•4 years ago
|
||
I did a couple of rounds of this on my local machine already. I fixed a few issues and then hit two bugs in symbolic. I think we should wait to do more here until after the symbolic bugs are fixed. Making this block on those.
|
|
Reporter | |
Comment 2•3 years ago
|
||
There's a bunch of scripts in the systemtests/ directory for building stacks based on crash reports in Socorro and symbolicating those stacks.
For v5, we expect these differences:
- Eliot response will have "line" in the output, but Tecken response will not
- Eliot response will have a correct "module" for each frame
- Eliot will have a "file" in the output where appropriate
I've looked at a few hundred stacks now and things look ok.
Marking as FIXED.
|
|
Reporter | |
Comment 3•2 years ago
|
||
Moving to Eliot product.
Description
•