When I try to modify either radius 1 or diameter 1 by typing in a new value, after I press enter, half of the time a random new value is used which is very close to the one I typed.
That can happen, because WinFIG has a discrete precision of 1200 dpi. That means many values must be rounded. For instance a diameter of 1 cm is changed to 0,9991 cm.
The reason is: 1 inch has 1200 dots in WinFIG, so 1cm has 1200/2.54 = 472.4409449 dots, which must be rounded to 472, because WinFIG stores coordinates in discrete values.
When you call the dialog the next time WinFIG makes the reverse calculation from dots to cm: (472 / 1200) * 2,54 = 0,9990666666, which is rounded to 0,9991 cm
The problem should occur less when working with imperial coordinates, but it can still happen, i.e. 0.9123in is 1094.76 points and would be rounded to 1095 points, which again displays as 0.9125in.
0.9123 * 1200 = 1094.76
1095 / 1200 = 0.9125
Sometimes I have to modify the value 3 times before the random "error" goes away and I end up with the precise value I wanted.
Yes, I have seen that there are rounding errors that occur even when a coordinate translates into a non-fractional number in a 1200 dpi system. I will try to find out what the reason is. This is always tricky due to the mixture of discrete coordinates and floating point calculations.
2. The "UPDATE" button on the main toolbar does not seem to apply to circles.
This is something I cannot replicate. It works fine here. If you mean the radius is not copied, that's right. You can find the attributes being copied int the list of check boxes on the right tool bar.
This means that when I want to scale a compound object, but keep the radii of the circles fixed, I have to manually go back and modify the "Values..." menu for each circle in the compound object. In conjunction with bug (1) this takes a lot of time to do.
That means you want to scale a compound but not the circles inside it?
While fixing bug (2) would help, I can't help observing that maybe an option to scale only (x,y) coordinates would also be useful (keeping other attributes fixed, e.g. radius).
ok, yes, that is something I could implement. The scaling would then just change the distribution of the circles.