FAIL: test_der (__main__.X509NameTests) / FAIL: test_digest (__main__.X509Tests)

[sandrotosi]

When building pyopenssl 0.14 on Debian (with python-cryptography 0.5.2) I got these failures when running tests in a clean chroot:

FAIL: test_der (__main__.X509NameTests)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "OpenSSL/test/test_crypto.py", line 907, in test_der
    b('0\x1b1\x0b0\t\x06\x03U\x04\x06\x13\x02US'
AssertionError: b'0\x1b1\x0b0\t\x06\x03U\x04\x06\x13\x02US1\x0c0\n\x06\x03U\x04\x03\x0c\x03foo' != b'0\x1b1\x0b0\t\x06\x03U\x04\x06\x13\x02US1\x0c0\n\x06\x03U\x04\x03\x13\x03foo'

======================================================================
FAIL: test_digest (__main__.X509Tests)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "OpenSSL/test/test_crypto.py", line 1454, in test_digest
    b("A8:EB:07:F8:53:25:0A:F2:56:05:C5:A5:C4:C4:C7:15"))
AssertionError: b'6E:DB:AA:35:8C:39:8C:86:95:0E:D6:51:7E:42:FF:F4' != b'A8:EB:07:F8:53:25:0A:F2:56:05:C5:A5:C4:C4:C7:15'

----------------------------------------------------------------------

the same on 2.7 and 3.4

[peterpramb]

For the first one see #115 - you must apply the fix there for OpenSSL >= 1.0.1h to let pyOpenSSL behave consistently.

[peterpramb]

The second one relates to #149 again - it's the same ASN1 structure change which causes different digest results.

OpenSSL 1.0.1e:

    0:d=0  hl=2 l=  37 cons: SEQUENCE          
    2:d=1  hl=2 l=  27 cons:  SEQUENCE          
    4:d=2  hl=2 l=   0 prim:   INTEGER           :00
    6:d=2  hl=2 l=   3 cons:   SEQUENCE          
    8:d=3  hl=2 l=   1 prim:    OBJECT            :itu-t
   11:d=2  hl=2 l=   0 cons:   SEQUENCE          
   13:d=2  hl=2 l=   4 cons:   SEQUENCE          

OpenSSL 1.0.1l:

    0:d=0  hl=2 l=  35 cons: SEQUENCE          
    2:d=1  hl=2 l=  26 cons:  SEQUENCE          
    4:d=2  hl=2 l=   1 prim:   INTEGER           :00
    7:d=2  hl=2 l=   2 cons:   SEQUENCE          
    9:d=3  hl=2 l=   0 prim:    OBJECT            :BAD OBJECT
   11:d=2  hl=2 l=   0 cons:   SEQUENCE          
   13:d=2  hl=2 l=   4 cons:   SEQUENCE          

[peterpramb]

Shortened test case from test_crypto.py:

from OpenSSL.crypto import X509, FILETYPE_PEM, dump_certificate

cert = X509()
cert.digest("MD5")
dump_certificate(FILETYPE_PEM, cert)

OpenSSL 1.0.1e:

cert.digest("MD5")
'A8:EB:07:F8:53:25:0A:F2:56:05:C5:A5:C4:C4:C7:15'
dump_certificate(FILETYPE_PEM, cert)
'-----BEGIN CERTIFICATE-----\nMCUwGwIAMAMGAQAwADAEHwAfADAAMAgwAwYBAAMBADADBgEAAwEA\n-----END CERTIFICATE-----\n'

OpenSSL 1.0.1l:

cert.digest("MD5")
'6E:DB:AA:35:8C:39:8C:86:95:0E:D6:51:7E:42:FF:F4'
dump_certificate(FILETYPE_PEM, cert)
'-----BEGIN CERTIFICATE-----\nMCMwGgIBADACBgAwADAEHwAfADAAMAcwAgYAAwEAMAIGAAMBAA==\n-----END CERTIFICATE-----\n'

[peterpramb]

Thanks, I can confirm that the test suite runs now cleanly with the fixes in #193 for both OpenSSL >= and < 1.0.1i.

[mrjefftang]

@peterpramb Are you seeing any test failures for set_tmp_ecdh() like those in the TravisCI runs?

[peterpramb]

No, I don't have this function at all in 0.14.

[peterpramb]

No issues using the latest master branch checkout with both OpenSSL versions:

DEBUG: test_set_tmp_ecdh (OpenSSL.test.test_ssl.ContextTests) ... ok
...
DEBUG: Ran 395 tests in 23.996s
DEBUG: OK

The only strange thing is that the test suite tries to build an extension:

DEBUG: running build_ext
DEBUG: building '_cffi__x6ebbce47xf0efe0ea' extension

[sandrotosi]

the test of the original post are still failing with pyopenssl 0.15 and python-cryptography 0.8.2-1 in the very same way

[alex]

This has been fixed for a while I believe.

[sandrotosi]

correct, this is fixed