Worklog, Encode Sans

[Stephen Nixon]

I have been doing some work to make Encode Sans into a variable font. This typeface was originally started by Pablo Impallari, and received extra attention from Jacques Le Bailly. 

Here's my GitHub repo: https://github.com/thundernixon/Encode-Sans

I will primarily document my work in the repo /docs, so you can check that source for the most up-to-date progress. However, I will also post periodic updates here for greater visibility in the community.

[Stephen Nixon]

I did some work to understand and get past a "weight cliff" issue. Thanks to the help of Mike LaGattuta, there is now a handy Glyphs Script to fix this.

So far, the script is a bit different for different fonts. Generally, however, it takes non-rectangular designspaces and makes them rectangular, so that sliders work across their ranges, and so that weight values line up across a second axis (e.g. Width).

For the full details of my work in deciding how to morph the designspace, see this documentation:

Variable Font issue: Variable fonts with non-rectangular designspaces export with unpredictable failures. FontMake (?) should fix this

There is another, somewhat-associated issue: kerning groups get confused during the FontMake export, yielding some very odd kerning. This issue is documented at https://github.com/googlei18n/fontmake/issues/470.

Here's an overlay showing the current variable font overlaid onto the existing static version. This problem gets worse with increased font weight, because the positive kerning values for /dcroat are being mapped onto glyphs in the same group as /d (all lowercase glyphs with a straight right side).

[Stephen Nixon]

Font Baking Encode Sans

After getting past the "weight cliff" issue (see previous doc), I've had more work to do in getting past errors and failures flagged by FontBakery.

Font-mastering techniques I've found useful

Using a build script

I have had quite a bit of success in correcting issues by using a build script based on a shell script inherited from @mjlagattuta (NOTE: currently only works in my Python3 environment). This script strings together useful functionality, allowing a few options based on some set variables:

• the GlyphsApp source is temporarily duplicated and get its designspace morphed (to fix the weight cliff issue)
• the FontMake variable font build is triggered
• the generated designspace is either kept or discarded
• gftools is run to fix various metadata errors
• ttx is used to generate an XML version of the VF so that:
  • the STAT table is replaced with a hand-edited STATpatch.xml
  • the NAME table is replaced with a hand-edited NAMEpatch.xml
  • in turn, I used a simple Python script to help myself accurately update numbers in snippets of XML
• ttx is used again to compile back to TTF
• the built fonts is moved into a timestamped folder
• FontBakery is run to QA output

I have kept a log of FontBakery runs as a GitHub issue.

Setting common weight values across widths

FAIL: The variable font 'wght' (Weight) axis coordinate must be 400 on the 'Regular' instance.

FAIL: The variable font 'wght' (Weight) axis coordinate must be 700 on the 'Bold' instance.

To fix this: Set the instances to a common weight value in the script which morphs the designspace to be rectangular and matches instance weight values. Before, the instances were slightly off. This meant that something with a weight of 85 would map to 700, but something with a weight of 86 (as was the case in the SemiCondensed regular instance) would map to 403.7037.

Installing fontvalidator and changing how FontBakery handles it

💔 ERROR: Checking with Microsoft Font Validator.
- com.google.fonts/check/037
- 💔 ERROR Mono runtime and/or Microsoft Font Validator are not available!
- 💔 ERROR Failed with FileNotFoundError: [Errno 2] No such file or directory: 'FontValidator.exe': 'FontValidator.exe'

This probably doesn't have a final solution, because fontvalidator seems to be undergoing some changes. However, I've gotten past this error and received QA from Font Validator by:

1. Installing the latest osX release from https://github.com/HinTak/Font-Validator/releases, then unzip
2. Change the unzipped binary's permissions, chmod 755 FontValidator
3. Moving the downloaded file into the same directory I use for my Python environment (for me, that's just ~/Environments/gfonts/bin).

Find general.py in the FontBakery code, then find lines 160–166. Change:

1. Change FontValidator.exe to FontValidator 2. Change + to -no-raster-tests




Update: the latest FontBakery just has this working, so now I've simply run pip3 install --upgrade fontbakery.

And now, I'm getting a bunch of FontValidator errors reporting! I'll next need to understand which to fix, and which to accept (assuming some might be fine).

FontVal errors

MS FontVal duplicates errors a lot, because it reports them for every glyph checked. I literally have 11,000+ lines of errors from it, but they boil down to these:

• 🔥 FAIL MS-FonVal: The version number is neither 0x00010000 nor 0x0001002 DETAILS: 0x00010003

A ton of these:

• 🔥 FAIL MS-FonVal: The device table's DeltaFormat value is invalid DETAILS: LookupList, Lookup[0], SubTable[0](PairPos, fmt 1), PairSet[0], PairValueRecord[0], Value1, XAdvDeviceTable, DeltaFormat = 32768-

• 🔥 FAIL MS-FonVal: Intersecting contours DETAILS: Glyph index 1

• 🔥 FAIL MS-FonVal: The Mac PostScript string does not match the Microsoft PostScript string DETAILS: mac postscript = EncodeSans-ThinCondensed, MS postscript = EncodeSans-Regular

• ℹ️ INFO Microsoft Font Validator returned an error code. Output follows :

• ℹ️ INFO MS-FonVal: The MarkSetCount of the GDEF table DETAILS: MarkGlyphSetsDef: MarkSetCount=3

• ℹ️ INFO MS-FonVal: Not an OpenType table, contents not validated

• ℹ️ INFO MS-FonVal: Not an OpenType table, contents not validated DETAILS: This table type is defined in the Apple TrueType spec.

A bunch of these:

• ℹ️ INFO MS-FonVal: Unable to perform test due to previously detected errors DETAILS: Glyph index 1 Test: ValidateSimpContMisor

Plus:

• ℹ️ INFO MS-FonVal: Loca references a zero-length entry in the glyf table DETAILS: Number of glyphs that are empty = 9

• ℹ️ INFO MS-FonVal: maxSizeOfInstructions via method #1 DETAILS: maxSizeOfInstructions=0, computed from the glyf table

• ℹ️ INFO MS-FonVal: No string for Typographic Family name (Name ID 16)

• ℹ️ INFO MS-FonVal: No string for Typographic Subfamily name (Name ID 17)

• ℹ️ INFO MS-FonVal: The post name has an unexpected value DETAILS: glyph = 1120, char = U+FFFF, name = uni0337

• ℹ️ INFO MS-FonVal: The post name isn't in uniXXXX or uXXXXX format and there is no Adobe Glyph List entry DETAILS: glyph = 1217, char = U+000D, name = CR

• ℹ️ INFO MS-FonVal: The post name isn't in uniXXXX or uXXXXX format and there is no Adobe Glyph List entry DETAILS: glyph = 1218, char = U+0000, name = NULL

• ℹ️ INFO MS-FonVal: Rasterization not selected for validation

More updates coming soon!

[Stephen Nixon]

To avoid sending a mega-email, I'm instead linking to documentation updates on my GitHub repo.

• I've continued to fix build issues. As a great update, the major kerning issue I was having has been fixed in FontTools, so the variable font is now building with proper kerning/spacing! https://github.com/thundernixon/Encode-Sans/tree/master/docs/03-fixing-build-issues
  
  
• I was able to work with Jaques de Baily to come up with good solutions for a few visual upgrades to letters like  /Germandbls , plus letters with slashes. https://github.com/thundernixon/Encode-Sans/tree/master/docs/04-visual-refinements
  
  
• I've fixed most of the FontBakery issues found in Encode Sans. As part of this, I've made a Python script to automatically abbreviate nameIDs 4 and 6 in a font's metadata, to prevent installation errors. https://github.com/thundernixon/Encode-Sans/tree/master/docs/05-working-through-fontbakery-issues
  
  
• I've worked to include static exports in the build process, in order to also keep the improvements in these backwards-compatible versions of the font. https://github.com/thundernixon/Encode-Sans/tree/master/docs/06-exporting-statics
  
  

Final steps for Encode will be:

• Splitting variable font into 5 separate width-based families, for use on Google Fonts before a width axis is supported
  
  
• Double-checking kerning in upgraded diacritics
  
  
• Running a FontLab6 test on glyphs, going through glyph-by-glyph to ensure that outlines are good (and fixing any other issues that might be flagged by FontLab)
  
  
• Making sure I'm doing the right thing for hinting (based on a conversation with Micah that will happen tomorrow)
  
  
• Test installing on Windows & Office
  
  
• Opening Pull Request with explanations on any remaining FontBakery fails and warns

[Dave Crossland]

Thanks Stephen! Sterling work :) 

On Mon, Nov 26, 2018 at 12:51 PM Stephen Nixon <ste...@thundernixon.com> wrote:

Final steps for Encode will be:

• Splitting variable font into 5 separate width-based families, for use on Google Fonts before a width axis is supported
  

Ah, I think the API wants just static fonts (for all design changes), then a VF font with a single Weight axis only for the normal width, and then a full VF :)