[FFmpeg-devel] [PATCH 2/2] mov: Don't export unknown/unhandled metadata types as if they were UTF8

[Martin Storsjö]

On Tue, 13 Jul 2021, Derek Buitenhuis wrote:

> On 4/1/2021 12:51 PM, Martin Storsjö wrote:
>> +        } else if (data_type > 1 && data_type != 4) {
>> +            // data_type can be 0 if not set at all above. data_type 1 means
>> +            // UTF8 and 4 means "UTF8 sort". For any other type (UTF16 or e.g.
>> +            // a picture), don't return it blindly in a string that is supposed
>> +            // to be UTF8 text.
>> +            av_log(c->fc, AV_LOG_WARNING, "Skipping unhandled metadata %s of type %d\n", key, data_type);
>> +            av_free(str);
>> +            return 0;
>
> Should we add any UTF-8 validation on our end too?

(Ah, here the reply arrived from yesterday)

I guess we could - but I see that as a separate thing to do which could be 
applied everywhere where we export metadata.

Here we have a flag that clearly identifies what kind of data it is 
(although we only recognize some of them), and if a type is set, which 
isn't utf8, we at least should bail out there.

For reference for myself and others, the types are defined here: 
https://developer.apple.com/library/archive/documentation/QuickTime/QTFF/Metadata/Metadata.html#//apple_ref/doc/uid/TP40000939-CH1-SW35

So 0 would mean unset/unknown/whatever, where we keep doing what we did 
before. But for other values, like utf16, various endian integers, bmps, 
whatever, just bail out.

// Martin