Re: mpost --parse-first-line

Reinhard Kotucha writes:
>>>>>> "David" == David Kastrup <[email protected]> writes:

>>> In my opinion the only reliable solution is to use the exec
>>> system call to run the required program.

>> It's used anyway.

> You mean the exec in mpost which executes tex, I suppose.
> The problem is in tex, not in mpost.

>> You probably mean that it should explicitly do

>> latex &latex

>> instead of

>> tex &latex

>> right?

> This would make sense if mpost would parse the first line of a
> verbatimtex/etex block but it doesn't look into it.  The contents of
> the block is copied to a temporary file and then tex is executed
> (which has to parse the first line of this file).

Note that the implementation (at least with web2c) isn't like that.

mpost finds a btex...etex or verbatimtex...etex and calls
- makempx, which calls
-- mpto to extract the tex-source
-- tex to typeset the tex-source
-- dvitomp to convert the dvi output to input usable by mpost

If the '%&-parsing' is to change in this sequence to call a different
engine, it can be done in makempx.

On unix systems, makempx is a shell script, which should make it easy
to play with changes in the way the input files are handled.  On
windows it is a binary.

You can use the TEX environment variable to pick the engine/default
format.  It's default is 'tex --parse-first-line'.


Provided all engines are given different suffixes to use for their
format files, you can check whether a format exists with kpsewhich (in
a script) of kpse_find_file (in a binary), then use the suffix to
determine a suitable engine to run it with.  If there's more than one
match we'd prefer the running engine, and otherwise it becomes a
toss-up.

-- 
Olaf Weber

               (This space left blank for technical reasons.)