Re: [AUCTeX-devel] This ascender bug in dvipng mode...

[Jan-Åke Larsson]

David Kastrup wrote:
> Well, actually I think that
>
> 7{currentfile token not{stop}if 65781.76 div
>
> should be a dead giveaway if you find a way to search for it
> efficiently.  It fishes 7 numeric tokens from the PostScript file
> itself and converts them into TeX units.  I think this has always been
> in this form in preview-latex.

Correct. And the code just after this snippet has changed. dvipng now
tests for "!userdict" followed by
"7{currentfile token not{stop}if 65781.76 div", in addition to
"!(preview version-string)" followed by "(tightpage)"

The previous test used a long string, which should match the \special
exactly. I suppose (given my minimal Postscript knowledge) that the
following might give the "right" Postscript:

   \special{!(}
   \special{!userdict begin/bop-hook{7{currentfile token not{stop}if 
65781.76 div DVImag mul}repeat 72 add 72 2 copy gt{exch}if 4 2 roll neg 
2 copy lt{exch}if dup 0 gt{pop 0 exch}{exch dup 0 lt{pop 0}if}ifelse 720 
add exch 720 add 3 1 roll 4{5 -1 roll add 4 1 roll}repeat <</PageSize[5 
-1 roll 6 index sub 5 -1 roll 5 index sub]/PageOffset[7 -2 roll [1 1 
dtransform exch]{0 ge{neg}if exch}forall]>>setpagedevice//bop-hook 
exec}bind def end}
   \special{!)pop}

But that depends on what dvips does to the specials, I guess. In any
case, the string in the middle special is the one used by dvipng<=1.5 to
identify tightpage mode. Any newlines in the string, generated by mail 
transfer should be single spaces.

I have prepared for the release, but I'll wait until I get a "go" from
you. I might want to add code to avoid "ignored header" messages for the 
surrounding "(...)pop" if you decide to use the above approach.

/JÅ