On 10/04/19 9:22 AM, Michal Hoftich wrote: >> Thank you very much. Does that also fix the >> \AtBeginDocument{\definecolor{...}} issue? Unfortunately the PGF/TikZ manual >> still doesn't build with TeX4ht, but that is not related to xcolor now. >> > > It should fix that. The main issue was the color name anyway. But I am > sure there are other issues with the manual. I've tried to convert it > using tex4ht two years ago and failed, mainly because the compilation > was too slow on my machine. I think some issues can be fixed using the > dvisvgm4ht driver: > > https://github.com/michal-h21/dvisvgm4ht > > It seems to produce much better results than the default driver. > Externalization should work too. I can help, but I am busy this week > with writing article for BachoTeX, next week should be better.
I would really appreciate that! The main difficulty I have is that TeX4ht hooks into all kinds of macros which results in altered expansion behaviour which blows up all over the place. For the xcolor problems I have found some workarounds, prior to your patches, which can be found on my tex4ht branch. https://github.com/hmenke/pgf/tree/tex4ht > > >> That is the first time I see this problem. Unfortunately, chemfig is an >> extremely complex package by itself, so I don't really know how those circles >> are drawn. It must have to with that because if I just draw circles with >> TikZ, >> no shift occurs. If it is possible for you to create a MWE without chemfig, >> you can report it on the PGF/TikZ bugtracker. > > I am sure that hunting for this bug would be quite difficult, it > really seems like a problem in chemfig. Hm, that might or might not be the case. Since the problem seems to be driver-dependent it indicates a PGF/TikZ problem, but I agree, it will be difficult to isolate. Cheers, Henri > > Best, > Michal >