This change makes sense to me

Thanks. Will wait a few days for @xia-chu.

I'm sure it is possible to run a binary for any arch (and probably endianness) on Linux with binfmt support.
I was compiling ARM and AARCH64 apps this way and even tricked Docker Hub to build ARM images on their infra, so maybe big-endian version of PPC would work too, though I'm too lazy to test it myself.

I ❤️ Linux: decided to give it a spin on s390x (notable BE platform) Ubuntu container under Docker.

Both before and after this patch following worker tests were failing identically:

-------------------------------------------------------------------------------
Scenario: RTCP Feedback PS RPSI parsing
  parse FeedbackPsRpsiPacket
-------------------------------------------------------------------------------
../test/src/RTC/RTCP/TestFeedbackPsRpsi.cpp:50
...............................................................................

../test/src/RTC/RTCP/TestFeedbackPsRpsi.cpp:40: FAILED:
  REQUIRE( item->GetPayloadType() == payloadType )
with expansion:
  2 == 1

-------------------------------------------------------------------------------
Scenario: RTCP Feedback PS VBCM parsing
  parse FeedbackPsVbcmPacket
-------------------------------------------------------------------------------
../test/src/RTC/RTCP/TestFeedbackPsVbcm.cpp:56
...............................................................................

../test/src/RTC/RTCP/TestFeedbackPsVbcm.cpp:46: FAILED:
  REQUIRE( item->GetPayloadType() == payloadType )
with expansion:
  2 == 1

===============================================================================
test cases:   39 |   37 passed | 2 failed
assertions: 1600 | 1598 passed | 2 failed

So none of existing tests cover this LE/BE distinction yet.

I then decided to test the actual media flow with Rust echo example, but again on both branches it neither crashed nor produces any audio or video on receiving side. I guess actual s390x hardware is needed to test it properly or I missed something important or it is not going to work on BE platform without more testing and fixes 🤷

Wow amazing catch! Hi @jmillan (once you are back), look at this!

Thank you for your work and excellent project.
When I was developing the zlmediakit project, I used part of the mediasoup code. Some users reported this problem when used zlmediakit.

This change makes sense to me

Thanks. Will wait a few days for @xia-chu.

The zlmediakit project extracts some code from mediasoup, but retains the source description and mediasoup copyright notice. I hope this won't make you unhappy.

This is where zlmediakit refers to the mediasoup source code

The zlmediakit project extracts some code from mediasoup, but retains the source description and mediasoup copyright notice. I hope this won't make you unhappy.

Actually you reported a real issue and provided all the needed info, so of course... thanks!