Go cgo "exec gcc: signal: killed" (OOM) - Fix in CI

cgo spawns gcc/clang to compile the C side of a package. A large translation unit on a small runner can exhaust RAM, and the kernel OOM killer sends SIGKILL - surfacing as "signal: killed".

What this error means

A cgo build fails with exec: "gcc": signal: killed or the compiler dies with no diagnostic. Memory pressure on a small runner caused the kernel to kill the compiler process.

# example.com/app/native
cgo: C compiler "gcc" failed: exit status 1
exec: "gcc": signal: killed

Common causes

Runner ran out of memory

A heavy C compilation exceeded available RAM and the OOM killer terminated gcc with SIGKILL.

Too much build parallelism

Many gcc processes ran at once via -p, multiplying peak memory beyond what the runner has.

How to fix it

Lower build parallelism

Cap the number of concurrent compile actions so peak memory stays bounded.

shell

go build -p 2 ./...

Build on a larger runner

Move the cgo-heavy job to a runner with more RAM so the compiler is not killed.

.github/workflows/ci.yml

jobs:
  build:
    runs-on: latchkey-large

How to prevent it

Size cgo-heavy jobs onto runners with adequate RAM.
Cap -p so concurrent gcc processes do not multiply memory use.
Watch peak memory of native builds and split large C units.