@DavidCarlisle I think I can't build hyperref - the script seem to assume that the main folder is called "hyperref" and on my system it is called "ho-tex-hyperref" ;-)
@DavidCarlisle really strange, I reverted it back to newcommand, and added the \let part to the headers as well. Then it works. Probably something with how \markboth expands things
I have a four thread GNU parallel running the oberdiek dtx files right now. Let's see how long this takes on my home PC (which is also old, and does not have the same RAM as my work PC)
@daleif with newcommand the gobble will happen first so nothing in the header, with robust command then the command will be placed in the header so the definition needs to be in scope there
@DavidCarlisle Then I have no idea why this works. I removed the robust definition for \gobbleone and it keeps throughing errors in the header. When I add the let to the header definition all is well.
@UlrikeFischer isn't the problem more related to babel redefining \markboth, injecting \foreignlanguage into it, and making everything token lists instead. Thus the redefinition runs out of scope.
@DavidCarlisle assuming everything compiled correctly. Your bash look, running in four parallel threads:
@UlrikeFischer I know, I think most of the custom macros can be ignored. As soon as any macro with \gobbleone is hit in the index via `\idxmark{...} then it goes wrong. Maybe I can do something ultra short. One moment.
@UlrikeFischer super simple:
\documentclass{book} %\usepackage[english]{babel} % dies if babel is used \newcommand\gobbleone[1]{} \newcommand{\idxmark}[1]{{\let\gobbleone\relax\markboth{\sffamily\bfseries#1}{\sffamily\bfseries#1}}#1} \begin{document} \idxmark{thing\gobbleone }, \end{document}
# make pdf rm *.bbl || echo no bbl for i in *.dtx do sem -j+4 --fg " lualatex ${i} if [ -f ${i/.dtx/.bcf} ]; then # biber ${i/.dtx} bibtex ${i/.dtx} fi lualatex ${i} if [ -f ${i/.dtx/.idx} ]; then makeindex -s gind.ist ${i/.dtx/.idx} lualatex ${i} fi "; done sem --wait
I just installed GNU parallel (there is a strange cite thing you have to go through the first time you use parallel/sem
@egreg then explain why it compiles without babel, and not with babel. It seems to be that because babel makes \markboth run with tokens, then the redefinition gets delayed until the header is actually executed. Whereas the normal version gets expanded right away?
@daleif No idea. By the way, if I add \usepackage{imakeidx} there is no error: apparently this is due to interactions with the juggles memoir does with the index.
