test: Add negative NaN test cases #475
Conversation
Codecov Report
@@ Coverage Diff @@
## master #475 +/- ##
=======================================
Coverage 99.54% 99.54%
=======================================
Files 54 54
Lines 16226 16265 +39
=======================================
+ Hits 16152 16191 +39
Misses 74 74 |
| EXPECT_THAT(execute(*instance, 0, {cnan, anan}), ArithmeticNaN(TypeParam{})); | ||
| EXPECT_THAT(execute(*instance, 0, {cnan, -anan}), ArithmeticNaN(TypeParam{})); | ||
| EXPECT_THAT(execute(*instance, 0, {-cnan, anan}), ArithmeticNaN(TypeParam{})); | ||
| EXPECT_THAT(execute(*instance, 0, {-cnan, -anan}), ArithmeticNaN(TypeParam{})); |
There was a problem hiding this comment.
Shouldn't we check the sign of the nan on the result side?
There was a problem hiding this comment.
That is not requires as any sign is fine.
There was a problem hiding this comment.
The spec restricts only the payload of the propagated NaN.
There was a problem hiding this comment.
The spec restricts only the payload of the propagated NaN.
That means it could be any sign?
There was a problem hiding this comment.
I mean in the above example, it is CPU specific whether -cnan, -anan will result in anan or -anan?
There was a problem hiding this comment.
Both cnan and -cnan are wasm canonical NaNs. Similarly for arithmetic NaNs. The sign is ignored.
There was a problem hiding this comment.
I don't know how CPUs propagate the sign bit, it probably depends on the opcode, too.
There was a problem hiding this comment.
I'd add a todo here to investigate this further.
No description provided.