@bot build-linux32 build-osx

@bot build-linux32 build-osx

This was a good question. There's multiple copies going on here and they aren't necessary.
While the handling of the outgoing data stream is not identical between the seeder and the
node software, the new WriteMessage function can easily write directly to the stream in
question instead of copying into an intermediate which then is copied into the outgoing stream.

• Improve the API for testing
• Use NetMsgType::GETADDR in place of a hardcoded string
• Other minor nits

• Rebase ontop of linker fixes: D8622
• Use netmagic from config instead of hardcoding the values in multiple places. (More cleanup can be done here, but out of scope of this patch)
• Use TransportSerializer.prepareForTransport to de-dup header serialization code.

There's an ava functional test, so at least original behavior still works as expected. Not much more to test here.

Functional test would be ideal. Manual test instructions in the test plan at minimum though. Both would be best :)

Fix clang build

Moved function to generated file and fixed test failure

Rebase

Rebase

Use the template on the clang builds. Even though changes to functional tests only will be a false positive,
catching changes to /cmake are important. Functional test only changes are not super common, so we
can live with this tradeoff to reduce maintenance overhead and reduce false negatives.

Used a template where it makes sense. For the others, it doesn't make sense to run them when
only changes to the functional tests are made.

Let's try this. If this doesn't work well, we can disable caching all together.

Please remove the reference to overrides.css from the bcha page too

Rebase to fix another issue. See D8482

Rebase with D8481 to fix CI

Rebasing again since buildbot did not finish re-deploying