COLR/CPAL requirement for glyph ID 1

[khaledhosny]

Fonts using COLR and CPAL tables must implement glyph ID 1 as the .null glyph.

The spec has the text above with no explanation why this is needed.

---

Document Details

⚠ Do not edit this section. It is required for docs.microsoft.com ➟ GitHub issue linking.

• ID: be879f46-c026-4a9d-1156-a3c98ae8f429
• Version Independent ID: 28be1ce2-e487-b652-632f-e4dc5bb17f42
• Content: COLR - Color Table - Typography
• Content Source: typographydocs/opentype/spec/colr.md
• Product: typography
• GitHub Login: @PeterCon
• Microsoft Alias: PeterCon

[bungeman]

It appears that in DirectWrite's current implementation (tested on Microsoft Windows Version 10.0.18363.657) when using a font with CFF outlines and COLR/CPAL the IDWriteColorGlyphRunEnumerator will always state that glyph ID 1 needs to be drawn first for all color glyph runs, as if every layer record started with an implicit gID of 1 (I haven't checked what the implicit paletteIndex is, I think it was either black or current color, though it may have been the paletteIndex of the actual first entry). This has come up in https://bugs.chromium.org/p/skia/issues/detail?id=9981 (call TranslateColorGlyphRun to get IDWriteColorGlyphRunEnumerator then MoveNext and GetCurrentRun to get the DWRITE_COLOR_GLYPH_RUN which has a DWRITE_GLYPH_RUN which always seems to want to draw glyph 1 the first time around).

I think this text may just be the result of this bug? I'll try to put together a simple reproducer if I can. Hopefully if this is fixed this text can be removed. If not, I'd be very interested in knowing why glyph ID 1 is being drawn under all color CFF glyphs and how that is supposed to work.

[khaledhosny]

Great finding, thanks. This then wouldn't be the only case where the COLR spec is modeled after the Windows implementation, another one would be #250.

[PeterCon]

I contacted the relevant developer at Microsoft asking him to provide input in this discussion. I had in mind for him to respond here, but instead he responded to me email with the following info:

The DWrite implementation used to try to combine multiple glyphs of the same color in the same glyph run, in order to reduce the total number of glyph runs rendered. When doing so, it might need to insert white space between non-consecutive glyphs to preserve spacing, and glyph ID 1 was used for this purpose. The requirement that glyph ID 1 be blank was written into the spec to allow for this implementation detail.

This changed a while back, but there are still systems out there that use the older implementation.

[behdad]

The DWrite implementation used to try to combine multiple glyphs of the same color in the same glyph run, in order to reduce the total number of glyph runs rendered. When doing so, it might need to insert white space between non-consecutive glyphs to preserve spacing, and glyph ID 1 was used for this purpose. The requirement that glyph ID 1 be blank was written into the spec to allow for this implementation detail.

That's exactly why specs should not be written by whoever makes an implementation....

[PeterCon]

The ask in this issue was for information, which has been provided. If there isn't any proposal for revision of the spec, it can be closed.

(One possibility might be to change the across-the-board requirement to a compatibility note; e.g., "Some early Windows implementations require...")

If there's no further discussion in the next few days, I'll close this.

[khaledhosny]

I know at least 4 non-MS implementations that do not check for this requirement, nor do they need any special handling of GID 1, so my suggestion is to change the spec and make it a note that old MS implementation had this requirement.

[PeterCon]

I know at least 4 non-MS implementations that do not check for this requirement, no do they need any special handling of GID 1

I know that Apple's implementation ignore this. I'll ping Microsoft with a draft revision to that effect, noting the situation in other implementations.

[PeterCon]

Draft revision:

Fonts using COLR and CPAL tables must implement glyph ID 1 as the .null glyph. If the COLR table is present in a font but no CPAL table exists, then the COLR table will not be supported for this font.

Note: Some early Windows implementations of the 'COLR' table require glyph ID 1 to be the .null glyph.

[PeterCon]

Reviewed by MS: they have no concerns with the proposed change.