Investigate flaky parallel/test-crypto-keygen

[Trott]

• Version: 11.0.0-pre (master)
• Platform: linuxone rhel72-s390x
• Subsystem: test crypto

https://ci.nodejs.org/job/node-test-commit-linuxone/5311/nodes=rhel72-s390x/console

09:59:09 not ok 2221 parallel/test-crypto-keygen
09:59:09   ---
09:59:09   duration_ms: 120.14
09:59:09   severity: fail
09:59:09   exitcode: -15
09:59:09   stack: |-
09:59:09     timeout
09:59:09     assert.js:662
09:59:09         throw actual;
09:59:09         ^
09:59:09     
09:59:09     Error: error:0D07207B:asn1 encoding routines:ASN1_get_object:header too long
09:59:09         at Sign.sign (internal/crypto/sig.js:83:27)
09:59:09         at testSignVerify (/data/iojs/build/workspace/node-test-commit-linuxone/nodes/rhel72-s390x/test/parallel/test-crypto-keygen.js:43:42)
09:59:09         at assert.throws (/data/iojs/build/workspace/node-test-commit-linuxone/nodes/rhel72-s390x/test/parallel/test-crypto-keygen.js:234:7)
09:59:09         at getActual (assert.js:574:5)
09:59:09         at Function.throws (assert.js:692:24)
09:59:09         at AsyncWrap.common.mustCall (/data/iojs/build/workspace/node-test-commit-linuxone/nodes/rhel72-s390x/test/parallel/test-crypto-keygen.js:233:12)
09:59:09         at AsyncWrap.<anonymous> (/data/iojs/build/workspace/node-test-commit-linuxone/nodes/rhel72-s390x/test/common/index.js:350:15)
09:59:09         at AsyncWrap.wrap.ondone (internal/crypto/keygen.js:42:14)
09:59:09   ...

[Trott]

Comment from @tniessen in #22660 (comment):

@Trott I fear there might be a small chance (0.392156862745098% if my calculation is correct) of OpenSSL successfully decrypting a key it is not supposed to decrypt, but leading to an invalid key and thus causing this error message. Let me try to confirm that before fixing the test.

[Trott]

While stress-testing locally with -j 96 --repeat 192, I got a different error:

=== release test-crypto-keygen ===                    
Path: parallel/test-crypto-keygen
assert.js:662
    throw actual;
    ^

Error: error:0D0680A8:asn1 encoding routines:asn1_check_tlen:wrong tag
    at Sign.sign (internal/crypto/sig.js:83:27)
    at testSignVerify (/Users/trott/io.js/test/parallel/test-crypto-keygen.js:43:42)
    at assert.throws (/Users/trott/io.js/test/parallel/test-crypto-keygen.js:234:7)
    at getActual (assert.js:574:5)
    at Function.throws (assert.js:692:24)
    at AsyncWrap.common.mustCall (/Users/trott/io.js/test/parallel/test-crypto-keygen.js:233:12)
    at AsyncWrap.<anonymous> (/Users/trott/io.js/test/common/index.js:350:15)
    at AsyncWrap.wrap.ondone (internal/crypto/keygen.js:42:14)
Command: out/Release/node /Users/trott/io.js/test/parallel/test-crypto-keygen.js
[02:08|% 100|+ 191|-   1]: Done 

[Trott]

Failed on a different platform in CI:

https://ci.nodejs.org/job/node-test-commit-arm/18620/nodes=centos7-arm64-gcc6/console

10:06:22 not ok 2388 parallel/test-crypto-keygen
10:06:22   ---
10:06:22   duration_ms: 240.71
10:06:22   severity: fail
10:06:22   exitcode: -15
10:06:22   stack: |-
10:06:22     timeout
10:06:22     assert.js:662
10:06:22         throw actual;
10:06:22         ^
10:06:22     
10:06:22     Error: error:0D07207B:asn1 encoding routines:ASN1_get_object:header too long
10:06:22         at Sign.sign (internal/crypto/sig.js:83:27)
10:06:22         at testSignVerify (/home/iojs/build/workspace/node-test-commit-arm/nodes/centos7-arm64-gcc6/test/parallel/test-crypto-keygen.js:43:42)
10:06:22         at assert.throws (/home/iojs/build/workspace/node-test-commit-arm/nodes/centos7-arm64-gcc6/test/parallel/test-crypto-keygen.js:171:7)
10:06:22         at getActual (assert.js:574:5)
10:06:22         at Function.throws (assert.js:692:24)
10:06:22         at AsyncWrap.common.mustCall (/home/iojs/build/workspace/node-test-commit-arm/nodes/centos7-arm64-gcc6/test/parallel/test-crypto-keygen.js:170:12)
10:06:22         at AsyncWrap.<anonymous> (/home/iojs/build/workspace/node-test-commit-arm/nodes/centos7-arm64-gcc6/test/common/index.js:350:15)
10:06:22         at AsyncWrap.wrap.ondone (internal/crypto/keygen.js:42:14)
10:06:22   ...

[tniessen]

@Trott This is most likely a randomly occuring problem with low probability (which obviously increases massively by being tested multiple times).

Changing /bad decrypt/ to /bad decrypt|asn1 encoding routines/ should be enough if my assumption is correct, I'll try to confirm that.

[mscdex]

There should be an opensslStack (or something similarly named) on the error message object that may help as well?

[tniessen]

Fixed in #22980, thanks for letting me know, @Trott. (Out of 25000 generated RSA keys, the test failed 98, which is close enough to the theoretical value to confirm my assumption about the decryption padding.)