[newbie] separating plot output from debug output

I'm attempting to refactor an R script that does a lot of plotting,
among other things. Ideally I'd like to do something like

setup ? ? # does pdf(...)
for each part of input {
?plot(process(part))
}
cleanup ? # does dev.off()

but have problems:

1 I'm plotting to PDF, so everytime I dev.off() creates a new file,
?and I want everything in one file (as does my boss :-)

2 I'm doing the work on a cluster, where I very much do not have root,
?and which has a fairly minimal set of installed packages, so I can't
?just call something external like 'pdftk' to merge the PDFs as I go.

3 As part of the processing, I printf status and debug messages, which
?I don't want in my PDF(s).

The solutions I can imagine are

1 Append to a single PDF, but I understand this is not feasible, no?

2 Create a buncha PDFs with code above, download them to my laptop,
?merge them to a single PDF, upload it. Feasible but annoying and
?kludgey.

3 Separate processing from plotting, e.g.,

setup ? ? # but not pdf(...)
for each part of input {
?write(process1(part), intermediate)
}
pdf(...)
for each part of intermediate {
?plot(process2(part))
}
cleanup ? # does dev.off()

?Again, feasible but kludgey.

4 No status and debug messages. I hope to be that good someday :-)

Am I missing something? Are there clean solutions to this problem?

TIA, Tom Roche <Tom_Roche at pobox.com>