[l2h] Re: includegraphics broken in latex2html-2002-2-1 ?

[Bruce Miller]

Herb W. Swan wrote:
> Bruce Miller <bruce.miller at nist.gov> wrote on Wed Jun  4 19:41:49 2003:
>>I downloaded the latest l2h (latex2html-2002-2-1, dated May 23,2003)
[...]
> I just downloaded the same version of LaTeX2HTML that you did, and I got the
> same result. 

The `same' is in same as me or same as before?
Ie. it works now, I hope!

> However, latex2html-2002-2-1 has a much more serious defect not
> present in latex2html-2002. 

Actually, with rotate=-90, I'm more surprised that latex got it right, than
that latex2html got it wrong! ... but seriously ... :>
In any case, graphicx is just passing it on to latex as traditional.

I installed an old copy of 2002 (downloaded on Mar 21, 2002) that I
had handy.  Both it and the current 2002-2-1 failed.

A clue!! 2002's dvips command was:
  /usr/bin/dvips -S1 -i  -E -o/tmp/l2h14262/image ./images.dvi
but 2002-02-1's was:
  /usr/bin/dvips -S1 -i  -Ppdf  -E -E -o/tmp/l2h15824/image ./images.dvi
Ross is getting more generous with the -E's!
Maybe your config went from 0 to 1, instead of 1 to 2?

The -E option tries to produce encapsulated postscript, and crops the
pages smaller, so the rest of the bitmap processing pipeline goes much faster
(especially for the typically small formula).

However, I suspect dvips is miscalculating the bounding box:
  %%BoundingBox: 133 719 675 1117
Without the -E it gives
  %%BoundingBox: 0 0 612 792
Since your figure is about 540x300 it looks like the actual graphs
got missed when -E was used.

If there's a "It should just work" solution, it's beyond me, maybe Ross
has some ideas. Two workaround solutions are:
 1) comment out any $DVIPSOPT = " -E" in l2hconf.pm
    (I assume it will run slower on more typical cases).
 2) Make your durn'ed figures rightsideup in the first place!!
    What're we? Australians!?!?!  Sheesh!