- check if the other fork handles this better
- submit test case upstream
- if this can't be addressed upstream, can we detect and work around this? (scaling fixes it after all, but is expensive)
Description
Description
Comment Actions
The newer zxing fork at https://github.com/nu-book/zxing-cpp is the solution to this:
- it can read all codes we previously failed on
- it's 30% faster and supports the "pure barcode" mode (ie. optimizing for reading from generated images rather than camera input)
- it doesn't leak memory on failed PDF417 reads
Problem though: its build system doesn't install anything, so it's not usable without patching.