Closed jondegenhardt closed 3 years ago
Merging #319 into master will increase coverage by
0.00%
. The diff coverage is100.00%
.
@@ Coverage Diff @@
## master #319 +/- ##
=======================================
Coverage 99.34% 99.34%
=======================================
Files 18 18
Lines 7068 7070 +2
=======================================
+ Hits 7022 7024 +2
Misses 46 46
Impacted Files | Coverage Δ | |
---|---|---|
common/src/tsv_utils/common/numerics.d | 99.21% <100.00%> (+<0.01%) |
:arrow_up: |
On at least some subsets the problems were a result of using 32 bit builds on Windows.
At present it comes down to two test cases that fail on Windows 64-bit builds. It's due to deltas in the underlying call to the
std.format.format
function between Windows and Unix/MacOs platforms. The calls(format("%.*f", 0, 0.6)
andformat("%.*f", 0, -0.6)
should return"1"
and"-1"
respectively, but instead return"0"
and"-0"
. The Unix and MacOS platforms return the correct results.Despite the discrepancy, this is a smaller set than was excluded under 32-bit builds. On 32-bit, it was appearing as if truncation rather than rounding was more widespread.
The formatNumber tests succeed for both DMD and LDC. However, LDC unit test builds are not enabled due to small differences in printed random values in
tsv-sample
tests.