cmd/link: TestMachOBuildVersion failures

[gopherbot]

#!watchflakes
default <- pkg == "cmd/link" && test == "TestMachOBuildVersion"

Issue created automatically to collect these failures.

Example (log):

=== RUN   TestMachOBuildVersion
=== PAUSE TestMachOBuildVersion
=== CONT  TestMachOBuildVersion
    link_test.go:473: [/home/swarming/.swarming/w/ir/x/w/goroot/bin/go build -toolexec /home/swarming/.swarming/w/ir/x/t/go-build3479085517/b3448/link.test -ldflags=-linkmode=internal -o /home/swarming/.swarming/w/ir/x/t/TestMachOBuildVersion1649661379/001/main /home/swarming/.swarming/w/ir/x/t/TestMachOBuildVersion1649661379/001/main.go]: exit status 1:
        runtime: /home/swarming/.swarming/w/ir/x/t/go-build3479085517/b3448/link.test: exit status 1
--- FAIL: TestMachOBuildVersion (14.52s)

— watchflakes

[gopherbot]

Found new dashboard test flakes for:

#!watchflakes
default <- pkg == "cmd/link" && test == "TestMachOBuildVersion"

2026-03-21 01:35 gotip-linux-ppc64le_power10 go@5f5f4ccd cmd/link.TestMachOBuildVersion (log)

=== RUN   TestMachOBuildVersion
=== PAUSE TestMachOBuildVersion
=== CONT  TestMachOBuildVersion
    link_test.go:473: [/home/swarming/.swarming/w/ir/x/w/goroot/bin/go build -toolexec /home/swarming/.swarming/w/ir/x/t/go-build3479085517/b3448/link.test -ldflags=-linkmode=internal -o /home/swarming/.swarming/w/ir/x/t/TestMachOBuildVersion1649661379/001/main /home/swarming/.swarming/w/ir/x/t/TestMachOBuildVersion1649661379/001/main.go]: exit status 1:
        runtime: /home/swarming/.swarming/w/ir/x/t/go-build3479085517/b3448/link.test: exit status 1
--- FAIL: TestMachOBuildVersion (14.52s)

— watchflakes

[dmitshur]

From triage, maybe this test doesn't need to run on GOARCH=ppc64le as well.

CC @golang/ppc64.

[gopherbot]

Change https://go.dev/cl/759260 mentions this issue: cmd/link: run TestMachOBuildVersion only on darwin

[pmur]

Is the ppc64le-power10 failure maybe an OOM failure? There shouldn't be any harm running this on ppc. The log has a suspicious second failure:

--- FAIL: TestLargeTextSectionSplitting (8.69s)
    ld_test.go:165: build failure: exit status 1
        cmd/go: /home/swarming/.swarming/w/ir/x/w/goroot/pkg/tool/linux_ppc64le/link: signal: killed
        
FAIL
FAIL	cmd/link/internal/ld	26.007s

@jkrishmys, would you be able to check the ppc64le-power10 dmesg log? That might give some hints.

[jkrishmys]

@pmur — thanks; the possibility of OOM appears to be the root cause.

I checked the logs on the OSU ppc64le Power10 builder host, and there is clear evidence of cgroup OOM kills (CONSTRAINT_MEMCG) in the relevant timeframe. For example:

Memory cgroup out of memory: Killed process ... (link)

In the same window, ld.test also invokes the OOM killer, and I see compile processes being killed as well. This is consistent with the observed CI failure:

cmd/go: .../link: signal: killed

This plausibly indicates that the failure is not a deterministic cmd/link issue, but rather the linker being killed by the kernel due to memory pressure within the builder container (limit ~13.6 GB).

I was not able to reproduce this locally on P10-ppc64le (both TestMachOBuildVersion and TestLargeTextSectionSplitting pass consistently with more headroom), which aligns with this being specific to the builder’s resource constraints.