Image conversion is now using the most performant available path (sse2 or avx2) even when the image width is not a multiple of 16 (sse2) or 32 (avx2)
Added code coverage through tarpaulin
Fixed bug in plane constraint computation
Check if the buffers contain enough data according to the desired conversion
Code was heavily refactored to keep into account for out of bounds errors. Previously this could result in undefined behaviour, since the library could access invalid data.
To ensure proper coverage, tarpaulin coverage was included. This was used also to fix 3. and 4.
Coverage reports now at 90%, but the actual coverage is even higher, because tarpaulin is not able to cover lines belonging to inlined functions or propagated constants.
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.
fabiosky
commented
3 years ago
Issue #10 , if available:
Description of changes:
Code was heavily refactored to keep into account for out of bounds errors. Previously this could result in undefined behaviour, since the library could access invalid data.
To ensure proper coverage, tarpaulin coverage was included. This was used also to fix 3. and 4. Coverage reports now at 90%, but the actual coverage is even higher, because tarpaulin is not able to cover lines belonging to inlined functions or propagated constants.
By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.