Open christophsax opened 1 month ago
What I switched too in a few packages is too .... preprocess the whole pdf vignette locally, include the generated (and compressed !!) pdf in the package and then use the trick by Mark van der Loo to have a minimal Rnw which ... just includes the pdf. That seems to avoid it because we then only operate on a (pre-)compressed pdf. I would have to check if I do that with other JSS papers too.
I think the key between what you do and what winbuilder does is that winbuilder may not default to the --compact-vignettes=gs+qpdf
argument so that it has something to compress left? I have not had an issue on submission.
Thanks!
You are suggesting I should ignore it and just submit and, if lucky, we don't have the issue there?
I talked to Uwe at the UseR and he has offered some help. If unlucky, he still may support me.
Preprocessing would be plan B, I guess.
Yes. Especially if you recently spoke with him -- just sit the issue out, and I review. That cannot possibly be a real blocker. I would have to count them but I probably have a number of packages with pdf vignettes and I only do the pre-processing in some (mostly to avoid stumbling over esoteric fonts or styles; JSS should be fine at CRAN of course) so "wait and see" seems fine.
You could also just ask the above on r-package-devel, he may even chime un. Describe how on win-builder you get the warning, how building with the compression option locally does not and may ask/wonder if winbuilder does not compress by default or if there is another trick.
Will submit once CRAN is back from vacation.
On Winbuilder, I get the following WARNING:
I cannot reproduce it and don't understand what Winbuilder is doing.
I tried the following:
R CMD build --compact-vignettes=gs+qpdf on Mac, Windows, and Linux. The resulting file is approximately 600Kb everywhere, same as the input.
tools::compactPDF("seas.pdf", gs_quality = "ebook", verbose = TRUE)
==> (new=2.11485e+06)/(old=610372) = 3.46485 .. not worth using
tools::compactPDF("seas.pdf", gs_quality = "none", verbose = TRUE)
==> (new=609273)/(old=610372) = 0.998199 .. not worth using
Separately calling qpdf and gs (only on Mac) leads to no further reduction:
qpdf seas.pdf seas_compressed.pdf
-> 609Kbgs -sDEVICE=pdfwrite -dNOPAUSE -dBATCH -sOutputFile=seas_compressed.pdf seas.pdf
-> 2.2Mb