KDE & ARM (and introduction post to kde-core-devel)
kloecker at kde.org
Wed Jun 3 23:24:38 BST 2009
On Wednesday 03 June 2009, Michael Casadevall wrote:
> Sorry for the delay in responding to this thread, I've been
> traveling, then sick, then $REAL_LIFE all got in the way :-/. I
> currently have the build hardware that I'll be using for this task,
> but I probably won't have an opportunity to set it up until next
> week, or maybe two weeks from now due to real life
> 2009/5/31 Ingo Klöcker <kloecker at kde.org>:
> > On Sunday 31 May 2009, André Wöbbeking wrote:
> >> On Sunday 31 May 2009, Ingo Klöcker wrote:
> >> > qMax, qMin, qBound are template methods . So adding
> >> > overloads/specializations makes no sense and won't help.
> >> Why not? If you offer overloads/specializations for (double,
> >> float) and (float, double) for qreal == float platforms it should
> >> work.
> > Right, I didn't think about overloads taking mixed types. Those
> > overloads should only exist for platforms where qreal != double
> > because of the non-obvious loss of precision.
> This is a change that should likely be made in coordination with Qt
> upstream to see where they stand on the issue, and in the mean time,
> using the qMax<double>()
> (I think using double is better than qreal since its always safe to
> put a float in a double but not via versus).
That argument makes no sense because the result will anyway be stored in
a qreal (see Aaron's reply to the root of this thread). The whole point
of using qreal instead of float or double is that the most suitable
floating point type for the architecture to code is compiled for is
used. Also it doesn't make a difference whether 0.0 is interpreted as
float or as double (which is what was causing the problem; again, see
Aaron's reply to the root of this thread).
-------------- next part --------------
A non-text attachment was scrubbed...
Size: 197 bytes
Desc: This is a digitally signed message part.
More information about the kde-core-devel