regression(1.04) `nim doc main.nim` generates broken html (no css)

[timotheecour]

nim doc main.nim generates broken html (css file absent)
regression possibly introduced in #12816

Example

real life example: https://disruptek.github.io/gittyup/gittyup.html
minimal example:

proc fun*()=
  ## some proc
  discard

Current Output

Expected Output

as in nim 1.0.4:

Possible Solution

• either embed the css in html file or update instructions in nim --fullhelp for nim doc

Additional Information

D20191231T070930
chrome console shows:

nimdoc.css:1 Failed to load resource: net::ERR_FILE_NOT_FOUND
dochack.js:1 Failed to load resource: net::ERR_FILE_NOT_FOUND

it used to show:

dochack.js:1 Failed to load resource: net::ERR_FILE_NOT_FOUND

• working in version 1.0.4
• broken in b67dea7

[juancarlospaco]

nim rst2html also broken #12980 (comment)

Confirming broken doc gen with latest compiler build (as of ba4fbb6).

[timotheecour]

@yewpad the regression was likely introduced in #12816 , if you want to fix you can look at that PR :-)

So I got it working again. I had to put the CSS back into the nimdoc.cfg file so that the doc gen can actually include the css in the html file.

Problem is that @Araq wanted to have the CSS outside of the nimdoc.cfg file for either maintenance purposes or because of the tests failing when the CSS gets changed.

I can't exactly tell you why the doc command doesn't copy over the nimdoc.css file into the directory where the docs are being placed in but my guess is that the copyFile call only gets executed when koch docs is executed? Idk, it's hard to figure out. The doc gen implementation is so messed up and spaghetti-like that I didn't care anymore and put the css back in the config file.

Also, I don't consider the fix done just yet. I have to discuss with @Araq whether or not it is acceptable to have the CSS back in nimdoc.cfg. I think for now it will be fine but we'll see.

[timotheecour]

maybe open a draft PR and discuss it there?

[Araq]

Our koch tool copies the .js and .css files over. We need to move this logic inside the compiler. It's sad, I know.