x/crypto/openpgp/clearsign: TestMultiSign fails with "random source is broken" at Go tip

[dmitshur]

At Go tip, TestMultiSign fails when run without -short flag, including on first class ports like linux/amd64:

=== RUN   TestMultiSign
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
    clearsign_test.go:155: cannot create key: rsa: generated p == q, random source is broken
--- FAIL: TestMultiSign (6.93s)

(E.g., build 8729461560583552209.)

Though it started failing earlier, as of go.dev/cl/632477, with a different error ("clearsign_test.go:155: cannot create key: crypto/rsa: decryption error").

CC @FiloSottile, @rolandshoemaker.

[gabyhelp]

Related Issues

• x/crypto/openpgp/clearsign: TestMultiSign go test fails on armhf system #39831 (closed)

Related Code Changes

• crypto/rsa: fix keys with p < q

_{(Emoji vote if this was helpful or unhelpful; more detailed feedback welcome in this discussion.)}

[FiloSottile]

The test is indeed broken, and needs fixing.

[rolandshoemaker]

It's unclear if that test actually needs to generate deterministic keys? rand.Reader should just be fast enough for this use case, surely?

[gopherbot]

Change https://go.dev/cl/633655 mentions this issue: openpgp/clearsign: just use rand.Reader

[dmitshur]

Thanks for the quick fix! The monthly golang.org/x tagging workflow ran into this, and the fix enabled it to keep going (https://go.googlesource.com/crypto/+/refs/tags/v0.30.0).