The new PSfrag system no longer requires the ps2frag script; instead, it
handles the processing entirely within TeX/LaTeX and DVIPS. Im sure you
will agree that never needing to run ps2frag again is a nice convenience!
However, there are two significant differences in the way this new version
of PSfrag works. Please make yourself aware of them:
1) XDvi is no longer able to determine where your PSfrag replacements
go, so instead it lines them up in a vertical list to the left of the
figure. This allows you to confirm that they have been typeset properly, at
least. However, to confirm that PSfrag
positions your replacements properly, you will have to view the PostScript
version of your file with a viewer like GhostView, or print it out.
This seems to be the
only disadvantage to the elimination of the pre-processing step.
2) If you embed \tex commands inside your figures, you now need to
explicitly _tell_ PSfrag to process these commands. To do so, use
at the beginning of your LaTeX file.
If you only use \tex commands in a small number of figures, then a
more efficient might be to turn on \tex-scanning only for those
figures. To do that, add the command \psfragscanon immediately before
each relevant \includegraphics or \epsfbox command.