OTS is requiring data chunks in CMAP format 14 subtable in order?

[be5invis]

As read the source it looks like that, data chunks of Default UVS Tables and Non-Default UVS Tables must be ordered the same as varSelectors, but I did not see any requirements in the Spec restricting the order of data chunks.
If the data chunks are not ordered, OTS would report an error, triggered by line 531 in cmap.cc.

[khaledhosny]

PR with a test font, or just a test font, would be helpful.

[be5invis]

The first file having this problem is a Microsoft IP.
The second is here ↓
sarasa-gothic-j-regular.zip

Message:

ERROR: cmap: Bad subtable offset (206 != 27265)
ERROR: cmap: Failed to parse format 14 cmap subtable 2
ERROR at src\ots.cc:675 (`anonymous-namespace'::ProcessGeneric)
ERROR: cmap: Failed to parse table
Failed to sanitize file!

Format 14 subtable of this font:

        List of VarSelectorRecord #0 :  
        VarSelector:     fe00
        defaultUVSOffset:     0
        nondefaultUVSOffset:     26831

        List of VarSelectorRecord #1 :  
        VarSelector:     fe01
        defaultUVSOffset:     0
        nondefaultUVSOffset:     26812

        VarSelector:     e0100
        defaultUVSOffset:     8056
        nondefaultUVSOffset:     6877

        List of VarSelectorRecord #3 :  
        VarSelector:     e0101
        defaultUVSOffset:     6005
        nondefaultUVSOffset:     1066

        List of VarSelectorRecord #4 :  
        VarSelector:     e0102
        defaultUVSOffset:     1014
        nondefaultUVSOffset:     440

        List of VarSelectorRecord #5 :  
        VarSelector:     e0103
        defaultUVSOffset:     428
        nondefaultUVSOffset:     349

        List of VarSelectorRecord #6 :  
        VarSelector:     e0104
        defaultUVSOffset:     0
        nondefaultUVSOffset:     315

        List of VarSelectorRecord #7 :  
        VarSelector:     e0105
        defaultUVSOffset:     307
        nondefaultUVSOffset:     288

        List of VarSelectorRecord #8 :  
        VarSelector:     e0106
        defaultUVSOffset:     0
        nondefaultUVSOffset:     274

        List of VarSelectorRecord #9 :  
        VarSelector:     e0107
        defaultUVSOffset:     0
        nondefaultUVSOffset:     260

        List of VarSelectorRecord #10 :  
        VarSelector:     e0108
        defaultUVSOffset:     0
        nondefaultUVSOffset:     251

        List of VarSelectorRecord #11 :  
        VarSelector:     e0109
        defaultUVSOffset:     0
        nondefaultUVSOffset:     242

        List of VarSelectorRecord #12 :  
        VarSelector:     e010a
        defaultUVSOffset:     0
        nondefaultUVSOffset:     233

        List of VarSelectorRecord #13 :  
        VarSelector:     e010b
        defaultUVSOffset:     0
        nondefaultUVSOffset:     224

        List of VarSelectorRecord #14 :  
        VarSelector:     e010c
        defaultUVSOffset:     0
        nondefaultUVSOffset:     215

        List of VarSelectorRecord #15 :  
        VarSelector:     e010d
        defaultUVSOffset:     0
        nondefaultUVSOffset:     206

        List of VarSelectorRecord #16 :  
        VarSelector:     e010e
        defaultUVSOffset:     0
        nondefaultUVSOffset:     197

[be5invis]

This message is found here: https://github.com/khaledhosny/ots/blob/master/src/cmap.cc#L531
The code forces that the last UVS data block (DefaultUVSTable or NonDefaultUVSTable) must end at the end of the cmap subtable, which is not required by the OTSpec.

[khaledhosny]

If I drop this check, the santized file will be invalid, so something else seems to be broken as well.

[be5invis]

@khaledhosny What exactly do you mean "something"? Something in my font or something in your program?

[khaledhosny]

Something in OTS.

[extratype]

@be5invis I have the same problem with Source Han Sans fonts converted to ttf using otfcc-c2q. But you have passed OTS for Sarasa Gothic v0.5.4. How do you workaround this issue?

[be5invis]

TTX roundtrip can fix that...

[extratype]

Thank you, it works. Yesterday TTX just crashed somehow :p

[khaledhosny]

Fixed in #197