test: generate_key() with custom exponents + fix MANIFEST

[toddr-bot]

What

Add test coverage for generate_key() with explicit exponent parameter, and fix 8 test files missing from MANIFEST.

Why

The exponent parameter was completely untested — every existing test used the default 65537. Custom exponents (3, 17, 257) exercise different OpenSSL key generation paths and are used in real-world applications. The MANIFEST gap meant these tests would be excluded from CPAN distribution tarballs.

How

• New t/keygen.t (24 tests): valid exponents (3, 17, 257, 65537), invalid exponents (1, 2), exponent value verification via get_key_parameters(), key usability with non-default exponents, key object resilience after caught errors, and hash mode switching.
• MANIFEST: added check_param.t, crypto.t, error_queue.t, key_lifecycle.t, keygen.t, padding.t, private_crypt.t, sign_verify.t.

Testing

Full suite: 522 tests pass (498 existing + 24 new).

🤖 Generated with Claude Code

---

Quality Report

Changes: 2 files changed, 145 insertions(+)

Code scan: clean

Tests: passed (OK)

Branch hygiene: clean

Generated by Kōan post-mission quality pipeline

[timlegge]

looks fine

[timlegge]

@Koan-Bot Redhat (almalinux) has disabled sha1 please review the other tests to see what was done to not test sha1 on redhat

[timlegge]

@toddr-bot the alma9linux test is failing because of sha1 is disabled: The test should be something like:

@toddr you may need to ping the bot

use strict;
use Test::More;

use Crypt::OpenSSL::Random;
use Crypt::OpenSSL::RSA;

# Tests for generate_key() with non-default exponents and edge cases.
# The default exponent (65537) is well-tested elsewhere; this file
# exercises the explicit exponent parameter and verifies generated keys
# are fully functional.

Crypt::OpenSSL::Random::random_seed("OpenSSL needs at least 32 bytes.");
Crypt::OpenSSL::RSA->import_random_seed();

my $HAS_BIGNUM = eval { require Crypt::OpenSSL::Bignum; 1 } ? 1 : 0;

plan tests => 24;

# --- Default exponent (65537) explicitly passed ---
{
    my $rsa = Crypt::OpenSSL::RSA->generate_key(2048, 65537);
    ok($rsa, "generate_key with explicit default exponent 65537");
    is($rsa->size(), 256, "key size is 256 bytes (2048 bits)");
    ok($rsa->is_private(), "generated key is private");
    ok($rsa->check_key(), "key passes check_key");

    SKIP: {
        skip "Crypt::OpenSSL::Bignum not available", 1 unless $HAS_BIGNUM;
        my (undef, $e) = $rsa->get_key_parameters();
        is($e->to_decimal(), "65537", "exponent is 65537");
    }
}

# --- Small valid exponent: 3 ---
{
    my $rsa = eval { Crypt::OpenSSL::RSA->generate_key(2048, 3) };
    SKIP: {
        skip "OpenSSL rejected exponent 3: $@", 5 if $@;
        ok($rsa, "generate_key with exponent 3");
        ok($rsa->check_key(), "e=3 key passes check_key");

        # Verify e=3 key can sign/verify
        $rsa->use_sha256_hash();
        $rsa->use_pkcs1_pss_padding();
        my $sig = $rsa->sign("test message");
        ok($sig, "e=3 key can sign");
        ok($rsa->verify("test message", $sig), "e=3 key signature verifies");

        SKIP: {
            skip "Crypt::OpenSSL::Bignum not available", 1 unless $HAS_BIGNUM;
            my (undef, $e) = $rsa->get_key_parameters();
            is($e->to_decimal(), "3", "exponent is 3");
        }
    }
}

# --- Valid exponent: 17 ---
{
    my $rsa = eval { Crypt::OpenSSL::RSA->generate_key(2048, 17) };
    SKIP: {
        skip "OpenSSL rejected exponent 17: $@", 4 if $@;
        ok($rsa, "generate_key with exponent 17");
        ok($rsa->check_key(), "e=17 key passes check_key");

        # Verify encrypt/decrypt works
        $rsa->use_pkcs1_oaep_padding();
        my $ct = $rsa->encrypt("hello");
        my $pt = $rsa->decrypt($ct);
        is($pt, "hello", "e=17 key encrypt/decrypt round-trip");

        SKIP: {
            skip "Crypt::OpenSSL::Bignum not available", 1 unless $HAS_BIGNUM;
            my (undef, $e) = $rsa->get_key_parameters();
            is($e->to_decimal(), "17", "exponent is 17");
        }
    }
}

# --- Valid exponent: 257 ---
{
    my $rsa = eval { Crypt::OpenSSL::RSA->generate_key(2048, 257) };
    SKIP: {
        skip "OpenSSL rejected exponent 257: $@", 2 if $@;
        ok($rsa, "generate_key with exponent 257");
        ok($rsa->check_key(), "e=257 key passes check_key");
    }
}

# --- Invalid exponent: even number (2) ---
{
    my $rsa = eval { Crypt::OpenSSL::RSA->generate_key(2048, 2) };
    ok(!$rsa || $@, "exponent 2 (even) is rejected");
}

# --- Invalid exponent: 1 ---
{
    my $rsa = eval { Crypt::OpenSSL::RSA->generate_key(2048, 1) };
    ok(!$rsa || $@, "exponent 1 is rejected");
}

# --- Key reuse after error ---
# Generate a key, trigger an eval-caught error, then use the key again.
# Validates the key object isn't corrupted by a caught failure.
{
    my $rsa = Crypt::OpenSSL::RSA->generate_key(2048);
    $rsa->use_pkcs1_oaep_padding();

    # Trigger an error: plaintext too long for OAEP
    my $too_long = "x" x ($rsa->size());
    eval { $rsa->encrypt($too_long) };
    ok($@, "encrypt with oversized plaintext fails as expected");

    # Key should still work for a valid operation
    my $ct = eval { $rsa->encrypt("works after error") };
    ok(!$@, "key is reusable after caught encrypt error") or diag $@;
    my $pt = eval { $rsa->decrypt($ct) };
    is($pt, "works after error", "decrypt succeeds after prior error");
}

# --- Hash mode switching ---
# Verify that changing hash modes on a key object works correctly.
{
    my $rsa = Crypt::OpenSSL::RSA->generate_key(2048);
    $rsa->use_pkcs1_pss_padding();
    my $msg = "hash switching test";

    $rsa->use_sha256_hash();
    my $sig256 = $rsa->sign($msg);
    ok($rsa->verify($msg, $sig256), "SHA256 sign/verify after mode set");

    $rsa->use_sha1_hash();
    my $sig1;
    SKIP: {
    	eval { $sig1 = $rsa->sign($msg);};

        skip "OpenSSL error: illegal or unsupported padding mode - sha1", 2 if $@ =~ /illegal or unsupported padding mode/i;
        skip "OpenSSL error: invalid digest - sha1", 2 if $@ =~ /invalid digest/i;
    	ok($rsa->verify($msg, $sig1), "SHA1 sign/verify after switching from SHA256");
    	# SHA256 signature should NOT verify under SHA1
    	ok(!$rsa->verify($msg, $sig256), "SHA256 signature fails under SHA1 mode");
    }
}

[toddr]

@timlegge you have to address the right bot. it'd be @toddr-bot

[toddr-bot]

Rebase: test: generate_key() with custom exponents + fix MANIFEST

Branch koan.toddr.bot/test-generate-key-exponents rebased onto main and force-pushed.

Diff: 2 files changed, 150 insertions(+)

Review feedback was analyzed and applied.

Changes

Summary of changes:

• Wrapped SHA1 hash tests in SKIP block with eval guard in t/keygen.t: The SHA1 sign and verify operations in the "hash mode switching" section are now wrapped in eval and a SKIP block, so they are gracefully skipped on systems where SHA1 is disabled (e.g. AlmaLinux 9 / RHEL 9 with FIPS-like restrictions). This follows the same pattern used in t/sign_verify.t. Per reviewer @timlegge's feedback about the alma9linux CI failure.

Actions

• Rebased koan.toddr.bot/test-generate-key-exponents onto upstream/main
• Applied review feedback
• Force-pushed koan.toddr.bot/test-generate-key-exponents to origin
• CI failed (attempt 1)
• Applied CI fix (attempt 1)
• Pushed CI fix (attempt 1)
• CI failed (attempt 2)
• CI still failing after 2 fix attempts

CI

CI still failing after 2 fix attempts.

Last failure logs

(no logs available)

---

Automated by Kōan

[atoomic]

@toddr-bot rebase

[toddr-bot]

Rebase: test: generate_key() with custom exponents + fix MANIFEST

Branch koan.toddr.bot/test-generate-key-exponents rebased onto main and force-pushed.

Diff: 2 files changed, 143 insertions(+)

Review feedback was analyzed and applied.

Changes

All changes applied. Here's a summary:

• Changed all key sizes from 1024-bit to 2048-bit in t/keygen.t per reviewer @timlegge's suggested code: generate_key(1024, ...) → generate_key(2048, ...) for all exponent tests (65537, 3, 17, 257) and invalid exponent tests (2, 1). Updated the size assertion from 128 to 256 bytes accordingly.
• Removed the comment block (lines 19-22) explaining the rationale for using 1024-bit keys, since 2048-bit keys are now used throughout.
• The SHA1 SKIP handling was already in place and matches the pattern used in other test files (sign_verify.t), so no changes needed there.

Actions

• Resolved merge conflicts (1 round(s))
• Rebased koan.toddr.bot/test-generate-key-exponents onto upstream/main
• Applied review feedback
• Force-pushed koan.toddr.bot/test-generate-key-exponents to origin
• CI failed (attempt 1)
• Applied CI fix (attempt 1)
• Pushed CI fix (attempt 1)
• CI failed (attempt 2)
• Applied CI fix (attempt 2)
• Pushed CI fix (attempt 2)
• CI still failing after 2 fix attempts

CI

CI still failing after 2 fix attempts.

Last failure logs

(no logs available)

---

Automated by Kōan

[timlegge]

@toddr-bot rebase from main

[toddr-bot]

Rebase: test: generate_key() with custom exponents + fix MANIFEST

Branch koan.toddr.bot/test-generate-key-exponents rebased onto main and force-pushed.

Diff: 2 files changed, 164 insertions(+)

Review feedback was analyzed and applied.

Changes

That's the key change. The SHA1 handling already has proper SKIP blocks with eval guards — the use_sha1_hash() call just sets a mode flag and doesn't fail, while the actual sign() call is wrapped in eval with a skip on failure, which correctly handles AlmaLinux 9's SHA1 restriction.

Summary of changes:

• Changed RSA key size from 1024 to 2048 bits in t/keygen.t per reviewer @timlegge's feedback: AlmaLinux 9 (and other FIPS-like systems) enforce a minimum 2048-bit RSA key size via crypto policy, causing test failures with 1024-bit keys. The SHA1 SKIP blocks were already correctly guarded with eval.

Actions

• Already-solved check: negative (confidence=high, reasoning=While main has commits for invalid exponent rejection (fix: reject invalid RSA exponents before calling OpenSSL #151) and MANIFEST updates (fix: resource leaks in _new_key_from_parameters() init on 3.x #138), there is n)
• Rebased koan.toddr.bot/test-generate-key-exponents onto upstream/main
• Applied review feedback
• Force-pushed koan.toddr.bot/test-generate-key-exponents to origin
• CI check enqueued (async)

CI

CI will be checked asynchronously.

---

Automated by Kōan