Use tectonic to build the PDF of the manual

[ViralBShah]

Fix #45027, and also help #46704

Now that we have tectonic as an alternative in Documenter, and also as a JLL, it should be straightforward to use it in the PDF build of the manual. The main challenge is all the pygmentize stuff which seems to take forever.

I'm mainly opening this as a WIP PR. I do feel that this will make it easier for everyone to test the pdf build, if we can use the regular packages and artifacts to provide a tex installation.

@inkydragon Should this be sufficient to generate the PDF?

[ViralBShah]

The build fails with this:

TeX capacity exceeded, sorry [main memory size=5000000]

Do we need to do something like: https://tex.stackexchange.com/questions/7953/how-to-expand-texs-main-memory-size-pgfplots-memory-overload

[inkydragon]

The main challenge is all the pygmentize stuff which seems to take forever.

Not forever, but it takes a long long time to complete.

workaround: patch documenter.sty

 % listings
-\usepackage{listings, minted}
+\usepackage{listings}
+\usepackage[draft=true]{minted}  % without highlight

TeX capacity exceeded

maybe you just need:

tectonic -X compile --keep-intermediates --keep-logs -Z shell-escape  julia-doc-en.tex

[ViralBShah]

Thanks. That will help with my development, but not with CI, I imagine.

[ViralBShah]

Lots of unicode characters, mainly for emoji like things are missing in the fonts we use here.

Example:

warning: could not represent character "🧕" (0x1f9d5) in font "DejaVu Sans/OT:script=latn;language=dflt;mapping=tex-text;"
warning: TheJuliaLanguage.tex:40259: Missing character: There is no ���� (U+1F9D6) in font DejaVu Sans/OT:script=latn;language=dflt;mapping=tex-text;!
warning: could not represent character "🧖" (0x1f9d6) in font "DejaVu Sans/OT:script=latn;language=dflt;mapping=tex-text;"
warning: TheJuliaLanguage.tex:40259: Missing character: There is no ���� (U+1F9D7) in font DejaVu Sans/OT:script=latn;language=dflt;mapping=tex-text;!
warning: could not represent character "🧗" (0x1f9d7) in font "DejaVu Sans/OT:script=latn;language=dflt;mapping=tex-text;"
warning: TheJuliaLanguage.tex:40259: Missing character: There is no ���� (U+1F9D8) in font DejaVu Sans/OT:script=latn;language=dflt;mapping=tex-text;!
warning: could not represent character "🧘" (0x1f9d8) in font "DejaVu Sans/OT:script=latn;language=dflt;mapping=tex-text;"

[inkydragon]

The main challenge is all the pygmentize stuff which seems to take forever.

The main problem here is that we need to call the external program pygments in tex.
It looks like pygmentize can only be called one-at-a-time in tex, waiting for io most of the time.
Some parallelization is needed here.

some ideas:

• Try other highlighting packages: https://github.com/gpoore/pythontex
• hack minted package. Generate all the required highlighting cache files externally by hand, and then compile them with tex.

Lots of unicode characters, mainly for emoji like things are missing in the fonts we use here.

See JuliaDocs/Documenter.jl#1791
Support for displaying emoji is another issue that can be put aside for a while.

[ViralBShah]

How does texlive currently do it? I see the large compile times only with tectonic, I think.