Open CV copyTo function ( and probably other copying functions) looses precision! (or stops copying after a large number of values) (Bug #4431)

Description

[latest opencv built from source from github]
1) I read a binary file into an opencv Mat (lets say mat1).
2) I copy mat1 into another mat (lets say mat2), with no alteration whatsoever, with the copyTo function.
3) i write both mat1 and mat2 again into binaries.
4) I open both binary files in matlab (for its high precision) and do sum(sum(abs(mat1))) and sum(sum(abs(mat2))).
5) the two results are different by a significant amount !
6) In fact, i can see that the copying to the second mat seems to have "stopped" after some value. Why is that???
7) When the same copying is done by mapping both matrices as Eigen::Matrix and then copying mat1 to mat2 using the Eigen overloaded "+=+" operator, the result is 100% correct !

I would add the files, but they exceed 5GB in size. (working with big data).
    The files have values of double precision, from -1 to 1.
    I have to say that this issue is not reproduced for every such file, but happens only sometimes, like an undefined behaviour (although its consistent with this particular set of data).
    If you can allow me to submit this 5gb file, you will see for yourselves.

Can you, please, check whether the PR 4148 (https://github.com/Itseez/opencv/pull/4148) fixes this problem?

Maksim Shabunin wrote:

Can you, please, check whether the PR 4148 (https://github.com/Itseez/opencv/pull/4148) fixes this problem?

Will it be enough to replace only the changed file (modules/core/src/copy.cpp) , or do i have to download latest source from github and rebuilt?

The big contribution with IPP refactoring has been merged recently, so you will need to get full source snapshot: https://github.com/mshabunin/opencv/tree/fix-copyto-64

Also, please, note that there is a better solution for this problem being developed: https://github.com/Itseez/opencv/pull/4158

Turns out, i will have to wait at least until mid-july before i try anything, since i am on a time critical task. Sorry for the inconvenience.