CI: Test that we can produce artifacts with gcc

[oschaaf]

• First iteration on testing gcc support in CI. This restricts the
  task to just building Nighthawk, and not testing it yet, as that
  needs a change in Envoy to go in first.
• Small cleanup, and a bugfix for those who use the script to run
  locally, which would be detected as running in CircleCI.

Related issue: #382
(needs a follow up to close, left in as a code-level todo).

Signed-off-by: Otto van der Schaaf oschaaf@we-amp.com

---

• Prerequisite: Update Envoy to f6b86a58b264b46a57d71a9b3b0989b2969df408 #394

[oschaaf]

I plan to restart the CI tasks in here to get a sense of stability, as there are changes in here that impact memory requirements.

[oschaaf]

It looks like that when building the asan task with 3 concurrent jobs we get “Too long with no output (exceeded 10m0s): context deadline exceeded”fairly structurally instead of it being oom killed occasionally :( the link steps are very memory hungry, maybe we need to lower parallelism further. It might then also be worth looking into a way to wrap ‘ld’ to serialize just the link calls and have the compile steps run at higher concurrency, as that may save some time. That’s quite hacky though. Maybe there’s also a larger build VM available that we can use for asan. Memory seems very tight for asan; it may take a long time otherwise (approx 1.5-2hours).

[mum4k]

Thanks for adding this coverage.