Remaining steps needed for sign off

[m4rc1e]

I'm starting this thread so we can all agree on what needs to be done in order to release this family.

Personally, I think we have to do the following:

Diff fonts created from both designspaces and keep Roboto-min.designspace if diffs are acceptable If both designspaces create fonts which are acceptable, we should keep just the Roboto-min.designspace. The other designspace creates a VF using every .ufo (19 in total). The final file size is 3.6mb; the Roboto-min.designspace font is just 2.3mb.

Update project so no regressions occur when diffed against latest v2.138 static release When TN started this project, v2.136 was the latest release. We're now up to v2.138. The changes between the two releases is minimal. I think I just need to add the Condensed Medium instances, judging by the history, https://github.com/google/roboto/commits/master.

Make a final report which compares the VF against the v2.138 static ttfs In #4, I only diffed static ttfs generated with this chain against the v2.136 official release. I'm confident the chain can now output static ttfs which don't have any major regressions. We're now ready to compare the VF against the static unhinted instances. I've already started doing this

Merge this repo with the original google/roboto repo It would be great to just have one repo for Roboto. The original is able to output unhinted, hinted and webfonts, we should also have this. I guess we'll end up doing this later?

cc @davelab6 @anthrotype @cjdunn @sberlow

[davelab6]

On Fri, Jul 6, 2018, 3:07 AM Marc Foley notifications@github.com wrote:

I'm starting this thread so we can all agree on what needs to be done in order to release this family.

Personally, I think we have to do the following:

Diff fonts created from both designspaces and keep Roboto-min.designspace if diffs are acceptable If both designspaces create fonts which are acceptable, we should keep just the Roboto-min.designspace. The other designspace creates a VF using every .ufo (19 in total). The final file size is 3.6mb; the Roboto-min.designspace font is just 2.3mb.

I agree we should drop the first one and only have the -min one, if the output of that is acceptable. I believe that's what you had been resting and that it is.

Will you do this comparison, and will you remove the old unneeded files?

*Update project so no regressions occur when diffed against latest v2.138

static release* When TN started this project, v2.136 was the latest release. We're now up to v2.138. The changes between the two releases is minimal. I think I just need to add the Condensed Medium instances, judging by the history, https://github.com/google/roboto/commits/master.

Sounds good. Will you do this?

Make a final report which compares the VF against the v2.138 static ttfs

In #4 https://github.com/TypeNetwork/Roboto/issues/4, I only diffed static ttfs generated with this chain against the v2.136 official release. I'm confident the chain can now output static ttfs which don't have any major regressions. We're now ready to compare the VF against the static unhinted instances. I've already started doing this https://groups.google.com/d/msg/googlefonts-discuss/K2_FumuWLB8/c40M-YMDCQAJ

Great! This final report will be needed to proceed with the below

Merge this repo with the original google/roboto repo

It would be great to just have one repo for Roboto.

I agree.

The original is able to output unhinted, hinted and webfonts, we should

also have this. I guess we'll end up doing this later?

This is unhinted today.

While vtt has a VF autohinter that you've been using for GF Library families, I wonder if more can be done here. TTFAutohint VF support is also coming along... But regressing on the hinting from static to VF seems inevitable....

The webfonts are subsets of the hinted fonts, but I'll ask the GF engineering team if they want to continue doing that. Since currently VFs dump a whole family at you at once, but then you benefit from having the whole thing cached, it may not make sense any more to do that sunsetting.

[m4rc1e]

Sounds good, will you do this?

Yep, It's minor.

I'm currently working on v2.138 compatibility, the report and GFR/diffenator so we can have something acceptable for Cosimo. I'll post when I'm done

[davelab6]

Sounds good, thanks!

[davelab6]

@m4rc1e whats the status of this? :)

[m4rc1e]

We just need the medium instances.

@cjdunn did you attempt this? If not, I'll take a good stab at them on Monday.

[cjdunn]

@m4rc1e I did not. Please let us know how it goes on Monday. And if you need anything on my end I'd be happy to help as well, thanks.

[davelab6]

We just need the medium instances.

Those are now added with #22 being merged.

So, what's next?

[m4rc1e]

• tidy up repo (rename Roboto-min.designspace to Roboto.designspace) • bump version to v3.000

I'd like to cut a v3.100 version at some point later this year. That's when I'll combine this repo with the original. How does that sound?

[davelab6]

Perfect

[davelab6]

• tidy up repo (rename Roboto-min.designspace to Roboto.designspace) • bump version to v3.000

Who will do this, and when?

[davelab6]

Does VF version contain all codepoints defined in the older one?

Google engineers need a hard confirmation of this; @m4rc1e does fontdiff quickly tell us? :)

[m4rc1e]

Yep. It has the same amount of glyphs with same codepoints. Some glyph names may have changed though.

[davelab6]

Just FYI, Jill pointed out that some unencoded glyph names for the same glyph shape differ across styles in the pre-VF fonts, and that's been normalized.

[davelab6]

@m4rc1e please could you update this issue with the latest requirements. You mentioned on a call that the build isn't repeatable, for example.

[m4rc1e]

If we want this to be production ready so it can replace the original Roboto repo, we still need the following:

• STAT tables (use statmake)
• Ability to build, web, android sibling families.
• Repo cleanup. VTT hints should be stored as xml. There should be a single set of sources.
• Improve build chain so STAT and VTT hints are incorporated into final fonts. We may also want to add some post processing steps as well.

[davelab6]

@m4rc1e what is the issue with the OS/2 table? You were seeing different line heights on Chrome 69 or something?

[sannorozco]

Hallo @m4rc1e, I’ve managed to integrate vttlib into the build, but I’m having an issue with the compiling step, I’m getting this. KeyError: 'SOFFSET'

So I was wondering if you have a requirements.txt of your virtualenv, to compare with the one on this repo?

Have you encounter this error before?

Thanks!

[m4rc1e]

The 'SOFFSET' instruction is used for transformed components. gftools check-vtt-compile will tell you all the glyphs which vttLib cannot compile due to this issue. These glyphs should be decomposed then rehinted.

[sannorozco]

We now compile hints via vttLib, Thank you for your insight.