NaNpx values when font-size is set on the html element

[dotnetCarpenter]

I have posted this issue in cssnext#210 but feel that it really belongs here.

cat node_modules/pixrem/package.json | grep version: "version": "2.0.0",
node_modules/.bin/postcss -v: postcss version 5.0.4

Thanks to @ben-eb I have reduced a test case that reproduce the issue.
Running postcss/pixrem with the following command:
node_modules/.bin/postcss -u pixrem and/or node_modules/.bin/cssnext

Smaller test case:

html {
  font-size: 0.625vw;
}

h1 {
  font-size: 6rem;
}

Output:

html {
  font-size: 0.625vw;
}

h1 {
  font-size: NaNpx; /* only happens when font-size is set on the html element */
  font-size: 6rem;
}

[dotnetCarpenter]

Since you can't derive an absolute value from vh, vw, vmax, and vmin which are relative to the viewport, pre-runtime. I think pixrem should either throw a meaningful error or this must be a noop failing silently. I'm personal in favour of throwing, since it gives me an immediate response when preprocessing via either postcss -u pixrem or cssnext [--no-rem].

Note: by default the font-size of the root element (html) is 16px in all browsers on most systems. Even if you specify a percentage value, e.i. html { font-size: 62.5%; }, you can calculate the root font-size to be 10px.

[dotnetCarpenter]

Other values that might be problamatic is ex and ch, unless you can read the font being used in a meaningful way. Doable but not simple. This is however another issue, but one to be mindful of.

ex
This unit represents the x-height of the element's font. On fonts with the 'x' letter, this is generally the height of lowercase letters in the font; 1ex ≈ 0.5em in many fonts.
ch
This unit represents the width, or more precisely the advance measure, of the glyph '0' (zero, the Unicode character U+0030) in the element's font.

[iamvdo]

As Pixrem is one of the oldest PostCSS module, it needs to be rewritten to get better errors/warn messages, so for now, I'm in favor of failing silently to 16. All browsers that do support vw unit also support rem unit so this is not a big deal to fallback to 16 for older browsers.
Any opinion? cc @MoOx

[MoOx]

All browsers that do support vw unit also support rem unit so this is not a big deal to fallback to 16 for older browsers.

indeed.