| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Nobody really seems to care, they provide their own SSL bindings.
|
|
|
|
|
|
|
| |
Just changes to make the package pass tests. Some are just cosmetic
changes. Some would require proper investigation.
Fixes #310
|
| |
|
| |
|
| |
|
|
|
|
|
| |
With https://github.com/openssl/openssl/commit/2e06150e3928daa06d5ff70c32bffad8088ebe58
OpenSSL allowed verificaton to continue on UNABLE_TO_VERIFY_LEAF_SIGNATURE
|
|
|
|
|
| |
* Replace unittest.makeSuite with unittest.TestLoader().loadTestsFromTestCase
* import unittest from relative import
|
|
|
|
|
|
| |
Support of CentOS 6 ends on 2020-11-30.
Fixes #283
|
|
|
|
|
|
|
|
|
|
|
|
| |
The -cipher can not be used in OpenSSL 1.1.b+ for TLS1.3 cipher since
openssl upstream commit 1c31fe7eb093a ("Ignore cipher suites when
setting cipher list").
Use -ciphersuites for TLS1.3 cipher as documented.
Signed-off-by: Sebastian Andrzej Siewior <sebastian@breakpoint.cc>
Fixes #251
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
Linux rounds microseconds in the timeouts up to the HZ kernel parameter.
To avoid checking for rounded values, pick the timeout long enough
so that it is whole number of HZ for any reasonable HZ value.
By doing this, we also do not need to handle separately Windows
rounding microseconds down to milliseconds.
|
| |
|
|
|
|
|
|
|
| |
Specially replace complicated construct with unittest2 to one import to
M2Crypto top module.
Also, Rand.load_file should have first parameter as bytes(), not str().
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
For instance, without this patch no chunk decoding is performed, even
if the HTTP response includes a "Transfer-Encoding: chunked" header
(only affects the python3 code path). To fix this, "recv_into" has to
call http.client.HTTPResponse.readinto, which does the corresponding
transfer decoding.
Thanks to Marco Strigl <mstrigl@suse.com> for reporting and helping
to debug this issue.
|
| |
|
|
|
|
|
|
|
|
|
| |
platform.linux_distribution() has been deprecated for many versions, and
it will be removed in 3.7 (https://bugs.python.org/issue1322).
Also, we actually don’t need it (if we remove one unused SSLv2 test).
Fixes #205
|
|
|
|
|
|
|
|
|
|
| |
Without this change self.m2_ssl_free(self.ssl) is never called,
because m2.bio_noclose is defined as "0". Hence, the if-condition
is always false. This got broken in commit e2f707b172 ("SSL
package: Port to python3").
Note that these testcases rely on the "fact" (or CPython
implementation detail?) that "del s" calls s' __del__ method.
|
| |
|
|
|
|
|
| |
Make openssl_in_path into private property
Remove superfluous logging.
|
|
|
|
| |
And it makes output of testsuite run unreadable.
|
|
|
|
| |
It has been deprecated in py3k.
|
|
|
|
|
| |
* tries to fix HttpslibSSLSNIClientTestCase.test_IP_call() output buffer not
getting flushed. This caused an AssertionError.
|
|
|
|
|
|
| |
We don’t want to deal with IPv6 right here (e.g., Debian doesn’t point
'localhost' to :::1), and besides it is not supported on some minor
platforms (some claim OpenBSD).
|
| |
|
|
|
|
|
|
|
|
| |
On ppc64/ppc64le, test_HTTPSConnection_SNI_support has been failing with
> AssertionError: 'Hostname in TLS extension: "localhost"' not found in 'Setting secondary ctx parameters\nUsing default temp DH parameters\nACCEPT\n'
i.e. the connection-specific data has not been logged by s_server.
Wait before terminating it to allow it to flush its buffers.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
Using PyCharm "Inspect Code..." functionality I found plenty of
problems.
|
| |
|
|
|
|
|
|
|
|
|
| |
We have checked the version of OpenSSL only half way in the middle of
the test, but we started OpenSSL s_server always with -ssl2 paramter no
matter what version of OpenSSL we had. This time we are not running the
test at all, if the OpenSSL doesn't support SSLv2.
Fixes #145
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
(this was the original comment on removing util.py3{bytes,str}
describing my plans for future)
It seems to that this abstraction goes exactly wrong way. The
primary goal of py3k/py2k string abstraction should be IMHO the
elimination of the ambiguous py2k str() object. Everything should be
(using py2 terminology) either unicode() or bytes(), and when we
start working on rebasing the ``python3`` branch on top of ``six``,
we should spend all our effort on clarification between these two
types and elimination of str(). Which seems to me is exactly what
six.u() and six.b() methods do, so I would use them.
Also eliminate util.h2b and util.py3ord
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
I used the output of "2to3 -f import M2Crypto" to figure out
which imports needed to be fixed.
In addition, I added "from __future__ import absolute_import"
to all files that I touched. This brings in the new import behavior,
and works as far back as Python 2.5.0a1. See:
https://docs.python.org/3/library/__future__.html
https://www.python.org/dev/peps/pep-0328/
|
| |
|
|
|
|
|
|
| |
Some platforms don’t support them and others will follow.
Fixes #105
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
mitr: Doesn’t OpenSSL disable them by default nowadays? Not sure.
mcepl: Well, the question is for how old OpenSSL we want to develop
this. Some patches are originally from RHEL-5 …
mitr: It’s the other way around: the more versions we want to support,
the more tests we need to disable (or, better, autodetect the
precise support level, without relying on M2Crypto code, if
possible).
Blocked by #48
|
| |
|
| |
|
|
|
|
|
|
| |
For Python 2 compatibility, import print_function from the __future__ module.
This works as far back as Python 2.6.0a2:
https://docs.python.org/2/library/__future__.html
|
|
|
|
|
| |
This syntax was backported to Python 2.6:
https://docs.python.org/2/whatsnew/2.6.html#pep-3110-exception-handling-changes
|