Philip Taylor writes:
> >> If your two aux filenames are A and B, during the first run you read
> >> From A and write to B. Presuming no errors, during your next run
> >> however you need to read from B and write to A. How do you keep
> >> track?
> No idea how Frank does it, but at successful completion I close A and B,
> open B for input and A for output, and copy B to A in a robust environment
> (sorry, not the LaTeX sense of "robust" but then I don't speak LaTeX!).
> It _can_ go wrong during the copy, but only in the event of a machine failure
> or running out of disc space, provided that the environment is sufficiently
> robust. Since the final record copied can be (e.g.) "%! TeX copy complete",
> one can always search for this on re-opening to ensure that the file is valid.
that is more or less what i did implement as well if i remember
correctly. my idea was to provide for doing the copying within LaTeX
(the slow but portable variant) but also allow to set things up that
the copying is done by some OS script wrapper outside LaTeX instead.
i don't really think it is worth checking for diskfull problems as you
get some "can't rwrite on file xyz" error at the end of your previous
run anyway, but yes you could test for this. but if you do i would set
a tex executable flag at the end rather than trying to match some
string (also you would need to allow for the case when there is no
previous aux file at all, but this is all solvable)