\textext can fail in definitions in \startMPinclusions

[Sanjoy Mahajan <sanjoy@mrao.cam.ac.uk>]

Another test file for metafun.  I know that textext doesn't work in
loops.  I don't understand how it's implemented enough to know why, but
is the reason for its not working in loops the same as the reason it
doesn't work in included definitions (related to static analysis of the
code)?

The following file gives an example of what I mean.  It gives a page
saying "unknown", but if one uses draw \textext{directly} directly (the
commented-out line) then it works as expected.  I see the same behavior
with \sometxt{}.  [All with context 2006.08.08]

====================  cut here ====================
\starttext
\startMPinclusions
def yy = 
  draw \textext{via yy def}
enddef;
\stopMPinclusions

\startMPcode
  % draw \textext{directly};  % would give "via yy def" as the text
  yy;                         % gives "unknown" as the text on the page
\stopMPcode
\stoptext
====================  cut here ====================

Oh maybe here's the problem.  The log file says (1.tex is the test
file):

  system(mpost  -progname=metafun -mem=metafun 1-mpgraph)...executed.

Doing that by hand to get PDF gives a figure containing "unknown":

$ mpost  -progname=metafun -mem=metafun 1-mpgraph
$ mptopdf 1-mpgraph.1

But "texexec --mpgraphic 1.mp" produces the right figure.  So why does
\startMPcode run the wrong command?  The following works fine:

==================== cut here ====================
\starttext
\startMPinclusions
def yy = 
  draw \textext{via yy def};
enddef;
\stopMPinclusions

\startstaticMPfigure{a}
  yy;
\stopstaticMPfigure

\usestaticMPfigure[a]
\stoptext
==================== cut here ====================

The log shows startstaticMPfigure running texexec:

system(texmfstart --ifchanged=1a-a.mp texexec --mpgraphic 1a-a.mp)...executed.

Oh I think I know why mpost is being called with \startMPcode: It gets
fooled by the label hidden inside the def, so it thinks that there is no
need to run the more complicated texexec comment.  I tried two
hacks^H^H^H^Htests to check this:

1. Added \forceMPTEXgraphictrue right after \starttext, and then the
   figure came out fine.  Not sure whether this method will work if
   there is more than one figure (\ifforceMPTEXgraphic might get reset
   to false after each figure).

2. Added an explicit \textext{} in the figure:

    \startMPcode
      label(\textext{jj},(1in,1in));
      yy;
    \stopMPcode

  and the yy produces the right text.

Though I still don't know why the staticMPfigure doesn't need either
hack.

-Sanjoy