Jorgen Birkler comments

Repositories
Issues
Comments

Results 3 comments of


                                            Jorgen Birkler

Fix double serialization for very large exponents that loose precision with %f printf format flag to %g format flag

In test_io.cpp the write-read test cases fails for reading PI and a small number: ``` fabs((double)tl[0] - 0.0000000000001) >= DBL_EPSILON || fabs((double)tl[2] - CV_PI) >= DBL_EPSILON || ``` Basically with...

Fix double serialization for very large exponents that loose precision with %f printf format flag to %g format flag

Having some trouble compiling and run the tests. Ill try to contribute a updated test package for your review.

Fix double serialization for very large exponents that loose precision with %f printf format flag to %g format flag

Some more findings: If using %g instead of %.20g fails one test case in opencv. Not really sure why. In opencv they bitbang their own doubletostring: static char* icvDoubleToString( char\*...