[Libreoffice-bugs] [Bug 125099] Error in time calculations i version 6.2

[bugzilla-daemon at bugs.documentfoundation.org]

https://bugs.documentfoundation.org/show_bug.cgi?id=125099

Eike Rathke <erack at redhat.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|libreoffice-bugs at lists.free |erack at redhat.com
                   |desktop.org                 |

--- Comment #8 from Eike Rathke <erack at redhat.com> ---
> Eike: wouldn't just adding half of output resolution to the serial datetime
> be correct for proper rounding?
No, that is more or less what those commits were about, rounding up a date+time
or wall clock time into the next magnitude most times is wrong, especially if
23:59:59.9 would result in 00:00:00 then (together with date the next day).

The problem here is the underlying floating point value resulting from the
subtraction (if the format is cleared (Ctrl+M) on A1:A2 you see the floating
point values). Formatting with HH:MM:SS.000000 then shows the values
03:53:46.000002 and 03:23:59.999999 where the latter reveals the problem.

The correct way would be to format the result as duration, not wall clock time,
i.e. [HH]:MM:SS displays the expected 03:24:00 on A2.

However, users don't use that.. not sure if we can do anything about it, maybe
if for a formula the result format is automatically determined, but not if a
wall clock time format is manually applied. Or maybe if we can detect a
calculation result should fit into a certain wall clock time format.

Investigating.

-- 
You are receiving this mail because:
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/libreoffice-bugs/attachments/20190506/029b7dd2/attachment-0001.html>