test: Fix running spec tests that expect NaN result #484
Conversation
gumb0
force-pushed
the
spectest-nan
branch
4 times, most recently
from
afa6887 to
9e58854
Compare
Codecov Report
@@ Coverage Diff @@
## master #484 +/- ##
========================================
Coverage 99.56% 99.56%
========================================
Files 54 54
Lines 16858 17015 +157
========================================
+ Hits 16784 16941 +157
Misses 74 74 |
gumb0
force-pushed
the
spectest-nan
branch
2 times, most recently
from
a1c54c0 to
20beae1
Compare
test/spectests/spectests.cpp
Outdated
| const uint64_t uint_value = std::stoull(v.at("value").get<std::string>()); | ||
| /// The unsigned integer type matching the size of this floating-point type. | ||
| using UintType = std::conditional_t<std::is_same_v<T, float>, uint32_t, uint64_t>; |
There was a problem hiding this comment.
| using UintType = std::conditional_t<std::is_same_v<T, float>, uint32_t, uint64_t>; | |
| using UintType = FP<T>::UintType; |
| v.at("type").get<std::string>() == "f32" || v.at("type").get<std::string>() == "f64"); | ||
| assert(!is_canonical_nan(v) && !is_arithmetic_nan(v)); | ||
|
|
||
| // JSON tests have all values including floats serialized as 64-bit unsigned integers. | ||
| const uint64_t uint_value = std::stoull(v.at("value").get<std::string>()); |
There was a problem hiding this comment.
@gumb0 Did not mention before merging, because this question affected master too (e.g. not introduced via this PR): do we want to wrap any of these in proper exception handling for invalid inputs (see https://en.cppreference.com/w/cpp/string/basic_string/stoul for invalid_argument or out_of_range) or it is fine leaving it to the general exception catching we have?
There was a problem hiding this comment.
Actually it would be nice to output a line number in the case of exception, now you see only that exception happened somewhere in the file.
There was a problem hiding this comment.
Wait until it happens again.
No description provided.