com.google.fonts/test/034: xAvgCharWidth bakery vs glyphs vs font val

[m4rc1e]

When running check-google fonts on Barlow we get:

>> com.google.fonts/test/034 with ((u'font[16]', '/Users/marc/Documents/googlefonts/manual_font_cleaning/barlow/fonts/ttf/Barlow-Thin.ttf'),)
   Check if OS/2 xAvgCharWidth is correct.
 * FAIL: OS/2 xAvgCharWidth is 500 but should be 501 which corresponds to the average of all glyph widths in the font

Recently this test has been off by just a unit on the most recent versions of glyphs.

In Ms Val, it is a completely different story:

So we have:

Glyphs: 500
FB: 501
Ms Font Val: 420

In the MS Spec:

The value for xAvgCharWidth is calculated by obtaining the arithmetic average of the width of all non-zero width glyphs in the font. Furthermore, it is strongly recommended that implementers do not rely on this value for computing layout for lines of text, especially for cases where complex scripts are used.

https://www.microsoft.com/typography/otspec/os2.htm#acw

Against MS
When I look at the FB implementation, I can't see anything that would lead to such a drastic difference.

Against Glyphs
Perhaps they do floor rounding for this calculation?. We use round which will round higher any float >.5 e.g 4.5 is 5

cc @schriftgestalt

[anthrotype]

I wouldn't care too much about xAvgCharWidth. See @behdad's fonttools/fonttools#1079 (comment)

We discussed having an MVAR entry for that item but everyone agreed that value is useless...

[davelab6]

Maybe our check can accept values that off by 1-2-5 units?

[m4rc1e]

Could this also be set to a WARN?

[davelab6]

I would prefer the value to be somewhat sensical, since things like getflourish.github.io/anatomy-of-typefaces use it

[anthrotype]

yeah, WARN sounds good, and allow a tolerance of +/- 1 unit

[felipesanches]

OK. I think I have enough input to come up with a sensible solution for this.

I'm working on this one now.

[HinTak]

See HinTak/Font-Validator#33