Some canonical codepoints missing from Material Symbols 2.966

[henry-alakazhang]

We noticed in our font build that some of the canonical codepoints are missing from the latest built font files for Material Symbols. It seems like these were all recently migrated to new codepoints in June.

Specifically, these are no longer in the fonts:

• e674 (draft) / legacy e66d
• e677 (list_alt) / legacy e0ee
• e684 (movie) / legacy e02c
  (and their other names/codepoints)

The legacy codepoints are still in the font files but the new ones aren't. Our code currently references the canonical codepoints and they now show tofu.

[tphinney]

Dang. We have long had a situation where we had to support two (or more) duplicate icons with different names and did so by unifying them as a single glyph. We just had extra ligature support the extra name, both pointing at the same glyph, which got both codepoints.

We were not prepared for the possibility of future dis-unification of the formerly identical icons, when combined with the circumstance that one of them might be Google-only and the other public.

In short, that is what happened here. The code point exposed formally by Google on fonts.google.com has never changed. BUT, the glyph for that icon formerly supported an additional codepoint, which it no longer does. Anybody probing the font directly could get either or both codepoints.

The combination of factors that caused this problem has not come up previously in the history of Material Symbols, as far as I can figure. We have never dis-unified icons, let alone also had the misfortune that the unified icon(s) had a mix of public and internal Google-only icons. If the same combination of factors comes up again, I will flag it as a problem, and have us treat the variant icons as new rather than dis-unifying existing merged icons.

So, I can’t reasonably fix the existing problem (other stuff relies on the “missing” codepoints), but I hope to avoid any more of it in the future.

[henry-alakazhang]

Ah, I see, thanks for the response - that's unlucky. Ise674 now Google-only?

I see the latest release has restored draft e66d to the codepoint file variablefont/MaterialSymbolsOutlined[FILL,GRAD,opsz,wght].codepoints, which is fine for us since that's what we've reverted our code to use anyway. We use those files to validate our codepoints, so as long as they stay aligned, dis-unifications should also be relatively smooth.