Re: gdiplus: Add more tests for GdipMeasureString using device with custom resolution. Take 3.

[Dmitry Timoshkov]

Dmitry Timoshkov  wrote:
> This time with even more relaxed comparison of precalculated values (0.5 => 0.6).

To make things clearer:
I was working on this test under Windows with PSDK compiler, and once I replaced
division 1/72 by multiplication to 72.0 the calculated values started to almost
exactly match the ones returned by gdiplus APIs. Apparently testbot uses gcc
and that change didn't improve things much, so I had to relax the test to pass
under testbot.