TTFs not recognized as monospace by mintty in Cygwin

[markuspeloquin]

I cannot select the Roboto font for Cygwin. I found this, indicating the problem is that mintty cannot detect it as monospace:
http://superuser.com/questions/887300/how-to-install-a-font-onto-cygwin-terminal#answer-1075786

I'm not sure how mintty decides a font is monospace. I noticed that fontconfig's fc-query -f '%{spacing}' is different for RobotoMono. All fonts with values for spacing:

% for i in /c/Windows/Fonts/*.ttf; do
  spacing=$(fc-query -f '%{spacing}' $i)
  [[ -n $spacing ]] && echo "$(basename $i): $spacing"
done
consola.ttf: 100
consolab.ttf: 100
consolai.ttf: 100
consolaz.ttf: 100
cour.ttf: 100
courbd.ttf: 100
courbi.ttf: 100
couri.ttf: 100
DejaVuSansMono.ttf: 100
DejaVuSansMono_0.ttf: 100
DejaVuSansMono-Bold.ttf: 100
DejaVuSansMono-Bold_0.ttf: 100
DejaVuSansMono-BoldOblique.ttf: 100
DejaVuSansMono-BoldOblique_0.ttf: 100
DejaVuSansMono-Oblique.ttf: 100
DejaVuSansMono-Oblique_0.ttf: 100
DroidSansMono.ttf: 100
LiberationMono-Bold.ttf: 100
LiberationMono-Bold_0.ttf: 100
LiberationMono-BoldItalic.ttf: 100
LiberationMono-BoldItalic_0.ttf: 100
LiberationMono-Italic.ttf: 100
LiberationMono-Italic_0.ttf: 100
LiberationMono-Regular.ttf: 100
LiberationMono-Regular_0.ttf: 100
lucon.ttf: 100
marlett.ttf: 100
simsunb.ttf: 90
SourceCodePro-Black.ttf: 100
SourceCodePro-BlackIt.ttf: 100
SourceCodePro-Bold.ttf: 100
SourceCodePro-BoldIt.ttf: 100
SourceCodePro-ExtraLight.ttf: 100
SourceCodePro-ExtraLightIt.ttf: 100
SourceCodePro-It.ttf: 100
SourceCodePro-Light.ttf: 100
SourceCodePro-LightIt.ttf: 100
SourceCodePro-Medium.ttf: 100
SourceCodePro-MediumIt.ttf: 100
SourceCodePro-Regular.ttf: 100
SourceCodePro-Semibold.ttf: 100
SourceCodePro-SemiboldIt.ttf: 100

My guess is that the TTF contains at least one character with different spacing, so the TTF didn't get the monospace metadata.

[codeman38]

Yep, that's exactly the issue.

Looking at just RobotoMono-Regular.ttf:

• 1006 glyphs have an advance width of 1229 (the expected width).
• 7 glyphs have an advance width of 0 (this is also OK).
• 28 glyphs have some other width:

Width	Count	Glyphs
1106	6	acutecomb dotbelow gravecomb hook tildecomb uni030F
1259	5	Dcroat Eth dcroat hbar uni20AB
1249	4	Omegatonos Omicrontonos uni049E uni049F
1329	4	Epsilontonos Etatonos Iotatonos Upsilontonos
1230	3	f hyphen uni00AD
1379	2	dcaron lcaron
1227	1	at.alt
1269	1	tcaron
1449	1	ldot
2458	1	exclamdbl

(The letter "f" is 1 unit wider than other letters! This is definitely a bug...)

[codeman38]

Here's a slightly cleaned-up version of the script that I used to generate the above table:
https://gist.github.com/codeman38/b2aaf94fb18d688483e7bf93e436cfab

[davelab6]

@markuspeloquin and @codeman38 thanks for reporting and triaging this issue!

[laerm0]

@davelab6 I can hotfix this via ttx, but then I'll be doing unknown visual things to some of the glyphs. I opened it up in Glyphs to take a look, and a handful of glyphs – i.e. dcaron – are going to be outside the bounding box. I'm just going to fix them in Glyphs so I can nudge some outlines around if necessary.

[laerm0]

Discussion in #1832 suggests that fixing this will be a bad idea as it will likely break other sites using Roboto. Sorry for the slow response @markuspeloquin and @codeman38: best plan appears to be a fix/fork of your own.