[XeTeX] wrong paper size (geometry and truedimen)

[Jonathan Kew]

OK, I have checked in a correction to the xetex code, so that the pdf  
media size should be correct with the use of magnification.

However, there may be another issue as well: the positioning of the  
output on the paper, when magnification is used. The question is  
whether the default origin of the printed page, in the absence of  
\hoffset or \voffset, should stay at (1in, 1in) from the top left of  
the paper, or if this position should be subject to magnification. It  
is not clear to me what the "correct" TeX behavior would be, as this  
is not defined strictly by TeX but by output driver conventions. And  
in my experimentation, there seems to be a discrepancy between  
pdftex, on the one hand, and other paths such as tex+dvipdfm or tex 
+dvips+ghostscript on the other. XeTeX (understandably) behaves like  
tex+dvipdfm.

Maybe there are macro package issues that relate to this; I don't  
know how geometry.sty deals with the various output options, for  
example. Further investigation and suggestions are welcome.

JK