test: mark SEA tests as flaky on PowerPC

[joyeecheung]

Refs: #50740
Refs: nodejs/reliability#718

[nodejs-github-bot]

Review requested:

• @nodejs/single-executable

[richardlau]

Is this flaking/failing on main as well (I saw the reports of it crashing for the 20.10.0 proposal)?

[joyeecheung]

yes: nodejs/reliability#718

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/55665/

[richardlau]

I was curious that the updated doc also says the tests are skipped on s390x but they're not in the sequential.status file which is also updated in this PR -- it looks like for the single-executable-application tests there is a whole skipIfSingleExecutableIsNotSupported function:

node/test/common/sea.js

Lines 11 to 46 in 59b27d6

	function skipIfSingleExecutableIsNotSupported() {
	if (!process.config.variables.single_executable_application)
	common.skip('Single Executable Application support has been disabled.');
	if (!['darwin', 'win32', 'linux'].includes(process.platform))
	common.skip(`Unsupported platform ${process.platform}.`);
	if (process.platform === 'linux' && process.config.variables.is_debug === 1)
	common.skip('Running the resultant binary fails with `Couldn\'t read target executable"`.');
	if (process.config.variables.node_shared)
	common.skip('Running the resultant binary fails with ' +
	'`/home/iojs/node-tmp/.tmp.2366/sea: error while loading shared libraries: ' +
	'libnode.so.112: cannot open shared object file: No such file or directory`.');
	if (process.config.variables.icu_gyp_path === 'tools/icu/icu-system.gyp')
	common.skip('Running the resultant binary fails with ' +
	'`/home/iojs/node-tmp/.tmp.2379/sea: error while loading shared libraries: ' +
	'libicui18n.so.71: cannot open shared object file: No such file or directory`.');
	if (!process.config.variables.node_use_openssl || process.config.variables.node_shared_openssl)
	common.skip('Running the resultant binary fails with `Node.js is not compiled with OpenSSL crypto support`.');
	if (process.config.variables.want_separate_host_toolset !== 0)
	common.skip('Running the resultant binary fails with `Segmentation fault (core dumped)`.');
	if (process.platform === 'linux') {
	const osReleaseText = readFileSync('/etc/os-release', { encoding: 'utf-8' });
	const isAlpine = /^NAME="Alpine Linux"/m.test(osReleaseText);
	if (isAlpine) common.skip('Alpine Linux is not supported.');
	if (process.arch === 's390x') {
	common.skip('On s390x, postject fails with `memory access out of bounds`.');
	}
	}
	}

[joyeecheung]

Oh cool, I forgot about this helper, not sure what we should put in as the skip reasons though, maybe "On PowerPC, loading SEA results in SIGSEGV that's under investigation"?

[richardlau]

The SEA tests on PPC64 are not consistently crashing -- if they were https://ci.nodejs.org/job/node-test-commit-plinux/nodes=rhel8-ppc64le/buildTimeTrend would be more red. So should these be marked flaky rather than skipped?

[joyeecheung]

Switched to marking as flaky on PowerPC.

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/55685/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/55687/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/55688/

[nodejs-github-bot]

CI: https://ci.nodejs.org/job/node-test-pull-request/55709/

[mhdawson]

LGTM

[mhdawson]

Landed in 136742d

[joyeecheung]

It looks like we do need ppc64 instead of ppc https://ci.nodejs.org/job/node-test-commit-plinux/50941/