CIDSet, again

[a20god]

Here's a document that has a TrueType font with a CIDSet which contains 0x2000 octets of 0xff, that is, the first 65536 bits are set. How can this violate clause 6.2.11.4.2 of ISO 19005-2:2011? As CIDs must not exceed 65535, there cannot be a CID for which its bit in that CIDSet is not set.

6.2.11.4.2-todo-1.pdf

[bdoubrov]

It's been also quite some discussion at the Validation TWG on this CIDSet clause. The decision is that CIDSet shall identify exactly the glyphs present in the subset, but nothing else. So, in your example, CIDSet puts bit 1 to the glyphs which do not exist in the subset.

Note that "empty glyphs" in TrueType font are also considered as present, because, for example, these are all valid space glyphs.

[a20god]

That is, all bits from the first one (for CID 0) up to and including that for the highest CID that has a GID that is is less than or equal to the highest GID in the font must be one and all other bits coming after them must be zero. In other words, CIDSet could be semanticaly replaced by a single integer.

[bdoubrov]

In case of an embedded TrueType subset, I believe yes. Note that this is different between PDF/A-1 and PDF/A-2(3). In PDF/A-1 the only requirement is that this /CIDSet entry is present. In PDF/A-2 it becomes optional, but if it is present, it shall satisfy the above requirement.

[a20god]

For future reference: What I wrote above about CIDSet having to look like 11111...00000 (binary) is wrong as I (again) didn't take CIDToGIDMap into account (see https://github.com/veraPDF/veraPDF-validation-profiles/issues/147). Sorry.

[bdoubrov]

Looks like the latest version of veraPDF (1.16) does report this this file is valid. Need to recheck this issue.

[bdoubrov]

Looks like the issue is in the value of the BaseFont entry, which is different for the Type0 font in question and for the descendant CIDFont. Namely, Type0 font name does contain the subset prefix, but the CIDFont doesn't. So, veraPDF assumes that the CIDFont is not a subset and does not check the CIDSet entry.

To be discussed at the TWG.

[bdoubrov]

The subset prefix is only important for CIDFonts, as specified in Table 117 of PDF 1.7 specification. For Type0 fonts BaseFont does not play any special role as indicated in the Note of Table 121 of PDF 1.7 specification. So, verPDF behavior is correct in this case