CI to expose build instability

[rsheeter]

Expose #647. This check will fail until we fix that. I also moved build GS to be part of CI because in my mind it is.

[rsheeter]

Rebased on main

[anthrotype]

I believe it's still going to fail because of varying GSUB FeatureParams nameIDs

[anthrotype]

how about instead of just calling ttx -l you actually diff the output for two ttx -l calls? So it only shows what changed which is what we are interested in

$ diff -u <(pipx run fonttools ttx -l first-roman.ttf) <(pipx run fonttools ttx -l second-roman.ttf)
$ diff -u <(pipx run fonttools ttx -l first-italic.ttf) <(pipx run fonttools ttx -l second-italic.ttf)

[rsheeter]

I believe it's still going to fail because of varying GSUB FeatureParams nameIDs

That's right. I don't think we can merge it quite yet.

[anthrotype]

just rebased on main hoping that things just work

[anthrotype]

looks like gvar and head are still reporting same lengths but different checksums

https://github.com/googlefonts/fontc/actions/runs/7251757185/job/19754763563#step:12:19

[rsheeter]

looks like gvar and head are still reporting same lengths but different checksums

Guess we can't declare victory yet. Even so, I'm quite pleased GSUB is now fixed :)

[anthrotype]

After googlefonts/fontations#750 we can temporarily declare victory ✌️
We need to bump write-fonts first to see the fix but I can confirm that it now works locally (after exporting SOURCE_DATE_EPOCH of course).

[anthrotype]

with latest write-fonts 0.21.1, CI is now green!

[anthrotype]

shall I revert this until we figure out #647 (comment)? otherwise all new PRs are going to fail because of it..

[rsheeter]

I'd vote not, if CI is failing we shouldn't be merging much of anything except a fix and we still have the power to merge despite this failure if we really really want to