Re: [Denemo-devel] Integration test for 1.2.2

[Richard Shann]

On Sun, 2015-05-24 at 11:59 +0200, Andreas Schneider wrote:
> Am 23.05.2015 um 22:50 schrieb Richard Shann:
> > On Sat, 2015-05-23 at 21:27 +0100, Richard Shann wrote:
> >> (Though why they didn't pass for you when they passed
> >> for Travis I don't know). 
> > On second thoughts, this will be because the floating point unit on your
> > computer is rounding slightly differently from mine & Travis's ...
> 
> Maybe the reason was that I was building on an i386 installation. On my
> amd64 system, that error did not occur.
yes, it will depend on the very last bit in a floating point calculation
- we only store integer zoom values in the XML format because otherwise
we get involved in locale issues (23.5 or 23,5), so we store 100*zoom
and then divide afterwards. Unless it is rounded properly the result may
be one lower on some architectures/compilers.

Richard