store ctts table

[gBillal]

In this PR:

• We store the ctts table on muxing.
• Updated the first sample pts on demuxing, the first sample's dts is 0, however it's not necessary the same for the pts.
• Updated the test files (there's rounding errors due to conversion between the timescales, so the files are not the same after setting the add_dts_offset to false)

[mat-hek]

Hi @gBillal, thanks for the PR ;)

Updated the first sample pts on demuxing, the first sample's dts is 0,

I'm unsure if this is true. If the video PTS starts from 0, the DTS must start from a negative value, so that PTS is always greater than DTS, and that is required by, for example, FFmpeg libs.

[gBillal]

Hi @mat-hek
I'm not sure that I'm following you, that change to the first sample will always satisfy that condition (pts being always >= dts).

[mat-hek]

Hmm, I may have misunderstood it 🤔 let's approach it another way: why is this change needed: https://github.com/membraneframework/membrane_mp4_plugin/pull/115/files#diff-2d67f6484e673f781de79faf98a69c97f7bd56018e49d685b52b6678751c4fb0R53 ?

[gBillal]

Ah, you're talking about that. I did add that to check that the ctts table is not included if the diff between pts and dts is 0 for all samples. I was sure it'll be 0, because the streams doesn't include B-frames. However I didn't set the add_dts_offset to false for the hevc stream because it includes B-frames.

In other words, that change doesn't add much anyway, I can revert it to the old one.

[mat-hek]

@gBillal gotcha, so I'd revert that and add another test to check if ctts is absent when add_dts_offset is set to false. Otherwise looks good ;)

[varsill]

Seems good, thanks!

[mat-hek]

@gBillal please make sure that formatter, credo and dialyzer pass ;)

[gBillal]

@mat-hek what's the command you run for credo. mix credo works fine on my side

[mat-hek]

This is because of Elixir version and credo incompatibility. Please run mix deps.update credo or even better mix deps.update --all and it should be fine ;)

[varsill]

I believe @mat-hek 's approve is required before merging as he is the code owner of the plugin