Problem with f32x4.min, broken test runner
Categories
(Core :: JavaScript: WebAssembly, defect, P3)
Tracking
()
| Tracking | Status | |
|---|---|---|
| firefox81 | --- | fixed |
People
(Reporter: lth, Assigned: lth)
References
Details
Attachments
(1 external file)
There's a bug in ad-hack.js where the function 'eq' is defined twice with different semantics, thus removing a bunch of test cases from the float operations, notably including infinities and NaN. Fixing this by renaming the first definition as 'equal' reveals that there are test cases that are not passing, it looks like f32x4.min is the first to fail. That one needs to be fixed anyway, but let's first find out what's up here.
| Assignee | ||
Comment 1•6 years ago
|
||
For floating point, max and min are more complicated and we have to
use the appropriate logic. This bug was masked because of a different
bug in the test suite that prevented us from running NaN tests.
Depends on D85982
Updated•6 years ago
|
Comment 3•6 years ago
|
||
Backed out for SM bustages.
Failure log: https://treeherder.mozilla.org/logviewer.html#?job_id=312801749&repo=autoland
https://treeherder.mozilla.org/logviewer.html#?job_id=312801766&repo=autoland
Backout link: https://hg.mozilla.org/integration/autoland/rev/2ce43b1fe3db5a3121cb01e5ea3045da7b85b1c9
| Assignee | ||
Comment 4•6 years ago
|
||
A problem in bug 1656226, it needed to tweak some tests.
Comment 6•6 years ago
|
||
| bugherder | ||
Description
•