I reviewed this patch. I found these minor omissions which should be checked or implemented:

• Increase libmodulemd version
• Merge source code formatting commit
• Handle stream override in modulemd_build_config_equals()
• Handle stream override in modulemd_build_config_compare()

But I really don't like how it breaks modulemd_packager_v3_to_stream_v2() and modulemd_packager_v3_to_stream_v2_ext() (which you adjusted tests for):

The patch changes stream as a side effect in copy_packager_v3_buildconfig_to_stream_v2(). As a result modulemd_packager_v3_to_stream_v2() will assume a stream override of the latest buildconfig as a stream. It breaks modulemd_packager_v3_to_stream_v2() in the opposite direction than it's already broken in regard of buildopts. modulemd_packager_v3_to_stream_v2() adopts the first buildopts. And now it adopts the latest stream.

I understand that it stems from the ill-design of modulemd_packager_v3_to_stream_v2_ext() which produces a single modulemd-v2 object instead of a list the objects, an objects for each modulemd-packager-v3 buildconfig. I would swallow it as a necessary collateral damage. But I don't like that the damage is not in accord with buildopts. Could you please change the patch to:

• Either raise an error if the overrides differ among buildconfigs,
• or use an override of the first buildconfig as it is done with buildopts?
• Or ignore the override completely?

And document handling of the stream override in these conversion functions?