https://bugs.documentfoundation.org/show_bug.cgi?id=158190

Mike Kaganski <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

--- Comment #2 from Mike Kaganski <[email protected]> ---
As you correctly identified in https://gerrit.libreoffice.org/c/core/+/159193,
it is caused by a special handling of nDecPlaces == 0 with
rtl_math_RoundingMode_Corrected in rtl_math_round, when it doesn't use
rtl::math::approxFloor, as used with other values of nDecPlaces. This makes the
nDecPlaces == 0 ti actually not use "corrected" behavior, as defined in the API
[1]: 

> Like HalfUp, but corrects roundoff errors

The special casing of nDecPlaces == 0 with rtl_math_RoundingMode_Corrected
looks unneeded there at all, so as I suggested there in the change, the
preferred solution would be not fixing the special casing, but removing it, and
allowing the general case code to do its job.

Thank you!

[1]
https://opengrok.libreoffice.org/xref/core/include/rtl/math.h?r=b1fb6338&mo=2922&fi=108#106

-- 
You are receiving this mail because:
You are the assignee for the bug.

Reply via email to