The difference in diagonal measures isn't a very intuitive measure of error. You can get a better idea of the magnitude of the error if you know the angle in degrees, or even mm/m. The spreadsheet I wrote for aligning the axes in a printer it will convert the mismatch in the diagonal measure into degrees (and mm/m), then you can decide if it is close enough. The spreadsheet assumes matched lengths of the opposite sides of the "rectangle" that you're measuring- i.e. that it is actually a rectangle if the corners are at 90 degrees. If you can't cut at least two pieces to matched lengths, the spreadsheet isn't going to help. When you print an alignment cube (for which I wrote the spreadsheet) the printer is usually reliable when it comes to printing equal lengths- steps/mm in the X, Y, and Z axes shouldn't change during a print- if they do you have bigger problems than nonorthogonal axes!
The absolute squareness of the frame isn't critical- the squareness of the guide rails is. If you are designing a printer, design in a means (or at least think about how) to align the guide rails to be orthogonal and the frame squareness won't matter. It isn't necessary to make things adjustable the way the bed leveling is adjustable, with a thumb screw, for example. If you can use shims, that may be all you need to make a rail position as adjustable as it needs to be. In fact, guide rail alignment should be adjusted then locked down so it won't change because accurate alignment is not a fast, easy process. That is why I depise dual motor Z axis designs. It's just too easy for the X axis to get out of alignment.
If you want to make the frame square but can't mill the ends of the members, use flat plates with over-sized bolt holes to join them- that will allow some adjustability (but will give up some rigidity, unless you use a lot of plates and bolts).
Ultra MegaMax Dominator 3D printer: [
drmrehorst.blogspot.com]