deprecate old-style v1 JSON filter config

[mattklein123]

Now that TCP proxy has parity, we can completely remove v1
style config loading.

Risk Level: Low
Testing: New UT
Docs Changes: Added.
Release Notes: Added.

[mattklein123]

@alyssawilk @derekargueta PTAL

[derekargueta]

No code changes? Might be useful to users to hook this into the deprecated feature counter or start-up logging (although hopefully it's unlikely that anyone is still using deprecated_v1)

[derekargueta]

Also is this different than the v1 deprecation in the 1.8 deprecation notes (first bullet)?

[alyssawilk]

@PiotrSikora for parity comments, since this relates to #5355

If we're going to do this we should flag the actual API fields as well so the scripts pick it up, no?

[mattklein123]

If we're going to do this we should flag the actual API fields as well so the scripts pick it up, no?

Unfortunately there are no API fields here. This is related to this code:

envoy/source/extensions/filters/network/http_connection_manager/config.cc

Line 324 in 8d1ad35

if (filter_config->getBoolean("deprecated_v1", false)) {

and

envoy/source/server/listener_manager_impl.cc

Line 66 in 8d1ad35

if (filter_config->getBoolean("deprecated_v1", false)) {

I agree with @derekargueta that we need to engage the stat/warn logic here, but it will need special casing. I was being lazy. Let me look into how to handle this. @alyssawilk any quick thoughts on how to one off engage the logic here since we aren't dealing with proto fields?

[mattklein123]

@PiotrSikora for parity comments, since this relates to #5355

It doesn't actually relate to bind to port. See ^.

[alyssawilk]

Yeah, sorry, I was misreading the bit about the tcp proxy session and network filter config to think we were removing deprecated_v1 fields altogether

[mattklein123]

@alyssawilk I still want to address @derekargueta comment about engaging the stat/warn/fail by default behavior. Do you think it would be possible to actually guard these if checks with some type of fake runtime field check? I can look into it later today.

[mattklein123]

@derekargueta @alyssawilk updated PTAL

[mattklein123]

/retest

[repokitteh-read-only]

🔨 rebuilding ci/circleci: compile_time_options (failed build)

🐱

Caused by: a #7129 (comment) was created by @mattklein123.

see: more, trace.

[mattklein123]

@alyssawilk updated

[alyssawilk]

It makes me wonder if deprecatedFeatureEnabled should take an error argument, but I guess that'd bring us the same problem we have with configs where some call sites might otherwise get too spammy.

Also I guess I should go fix that TODO on deprecation tooling for features now :-)