possible math inconsistency

[AspirinJunkie]

Oh ok - i searched before but didn't found this ticket.
Anyway - the argumentation there is simply wrong.
It has nothing to do with a "max precision 15/16 digits" as it stated there.
There is no rounding needed because 2^49 can be mapped completely without rounding errors in double.
As jchd clearly explained, the rounding errors are to be expected only from 2^53. So here is a different problem.

Also, the internal type conversion in C/C++ does not show this problem.
So if it's possible to cast double to int64 without errors in C/C++ then it should be possible in AutoIt too.
I have explained all this there also so plausibly and think therefore that it is recognized this time correctly as a bug.
 

Edited April 8, 2021 by AspirinJunkie

[JockoDundee]

8 hours ago, AspirinJunkie said:

I have explained all this there also so plausibly and think therefore that it is recognized this time correctly as a bug.

Yes, however I think there is merit in referencing this ticket in your new ticket in so it doesn’t get marked as a duplicate.