Failing to correctly do AP4_CencSampleInfoTable::Create on particular file

[glennguy]

Hi again
The attached zip contains the init and a segment from an audio track of a CBCS encrypted stream:
audio.zip
This decrypts successfully in browser/official app, but bento4/Kodi/inputstream.adaptive is not currently handling this properly.

When trying to create a sampleinfotable it's missing data because there's no SAIZ/SAIO boxes, and the SENC box is effectively empty:

This here actually creates the sampleinfotable:

Bento4/Source/C++/Core/Ap4CommonEncryption.cpp

Lines 2828 to 2838 in 5b7cc25

	// try to create a sample info table from senc
	if (sample_info_table == NULL && sample_encryption_atom) {
	AP4_Result result = sample_encryption_atom->CreateSampleInfoTable(0,
	crypt_byte_block,
	skip_byte_block,
	per_sample_iv_size,
	constant_iv_size,
	constant_iv,
	sample_info_table);
	if (AP4_FAILED(result)) return result;
	}

But m_sampleCount is 0 (and probably more data missing too), no IVs are stored and ultimately calling DecryptSampleData fails here:

Bento4/Source/C++/Core/Ap4CommonEncryption.cpp

Line 2127 in 5b7cc25

if (iv == NULL) return AP4_ERROR_INVALID_FORMAT;

I'm hoping you can help with a solution or at least point me in the right direction. I've been at it for a couple of hours and tried a few things but now getting to the point of needing some help.

Thanks heaps.

edit: btw this is from the same stream/program with incorrect boxes in the video stream from issue #742

[glennguy]

Hi I have some more info.

So I now see that the 'data' in this box is indicating that there's a 0 count for sample info.

From this document: https://www.uvcentral.com/files/CFFMediaFormat-2_1.pdf on page 39

sample_count is the number of encrypted samples in this track fragment. This value SHALL be
either zero (‘0’) or the total number of samples in the track fragment.

I have a workaround here but not really sure if it's proper. I just copy/pasted code from another function that gets the sample info count from the trun box. With this number populated the table is successfully created. I also had to alter SetSampleInfoSize to update m_SampleInfoCount - this made sense to me: https://github.com/glennguy/Bento4-axiomatic/commit/5b5200b01a243fe4be7fd72559ac356ef3cc0cad https://github.com/glennguy/Bento4-axiomatic/commit/96c9abfc603be06310cb9902d87f1d12959b46ab

Grateful for your feedback.

[barbibulle]

This file is also not compliant (no saiz and saio, which are required by the spec).
I committed a change to the main branch that supports this case where saiz and saio are missing, and an senc box is present with sample_count == 0. In this case, all samples will use the same default IV.
Note that this is implemented with a special case for sample_count == 0 rather than building a table for all the samples, because just looking at the sample count in a trun box isn't enough to know how many samples there may be.
I couldn't verify that this works for your file, other than it not erroring out, because I don't have the decryption key, so I can't check that the decryption was actually correct.
Please let me know if this fix works for you.

[glennguy]

Thanks for looking at this @barbibulle
Works like a charm, thanks heaps for very fast turnaround!