exceptions: evaluate needed exceptions in upstream

[junr03]

This is largely a more speculative issue. According to analysis of sections within compilation units, most source files end up having a not insignificant size proportion in exception tables. @Reflejo recommended analyzing cases where exceptions might lead to irrecoverable conditions.

Size Win

per #17 (comment)

On the stripped binary there is also at least 1MB of exception tables. I don't know how many of them are exceptions we recover from vs we terminate. So we might investigate adding the nothrow specifier on methods so the except table is not created (not sure how we'll do this in an elegant way only for mobile, nor how we use exceptions, just throwing this here as a thought)

[mattklein123]

FWIW I would just skip tracking/looking into this. Every exception we throw today needs to be caught somewhere and I don't think we are going to reasonably be able to do anything major here. I think our time is better spent elsewhere.

[Reflejo]

Sounds like a dead end then. So if that's the case I'm ok not tracking/working on it

[alyssawilk]

coming back to this (thanks @Reflejo for the magic -fno-unwind-tables -fno-exceptions) it's still a pretty big gain to build without exceptions. From cce99bf8b0

-rwxr-x--- 1 alyssar primarygroup 18953104 Mar 14 09:40 bazel-bin/test/performance/test_binary_size_stripped
-rwxr-x--- 1 alyssar primarygroup 19M Mar 14 09:40 bazel-bin/test/performance/test_binary_size_stripped

-rwxr-x--- 1 alyssar primarygroup 17513048 Mar 14 09:37 bazel-bin/test/performance/test_binary_size_stripped
-rwxr-x--- 1 alyssar primarygroup 17M Mar 14 09:37 bazel-bin/test/performance/test_binary_size_stripped

[alyssawilk]

In my working PR I largely replace throw with throwEnvoyExceptionOrPanic which would be defined to PANIC for E-M

on start-up we already catch all exceptions and replace with panic. sans-XDS, any other exceptions would be on the data plane which is disallowed and theoretically doesn't happen.

The big gotcha here, is that jbeder_yaml_cpp also has exceptions which means loadFromYaml and friends don't work. Google was already hoping to move away from yaml for E-M since udp doens't support yaml loading, so I'd say that's a prereq for this reduction.

[alyssawilk]

running PR here: https://github.com/envoyproxy/envoy/compare/main...alyssawilk:envoy:size_exceptions?expand=1
down to ~90 files