* Always generate a new key pair if the private key doesn't exist (#597)
This commit updates `KeypairBackend._should_generate()` to first check
if the original private key named by the `path` argument exists, and
return True if it does not. This brings the code in line with
the documentation, which says that a new key will always be generated if
the key file doesn't already exist.
As an alternative to the approach implemented here, I also considered
only modifying the condition in the `fail` branch of the if statement,
but I thought that would not map as cleanly to the behavior specified in
the documentation, so doing it the way I did should make it easier to
check that the code is doing the right thing just by looking at it.
I also considered doing something to make the logic more similar to
`PrivateKeyBackend.needs_regeneration()` (the openssl version of this
functionality), because the two are supposed to be acting the same way,
but I thought that'd be going beyond the scope of just fixing this bug.
If it'd be useful to make both methods work the same way, someone can
refactor the code in a future commit.
* Test different regenerate values with nonexistent keys
This commit changes the test task that generates new keys to use each of
the different values for the `regenerate` argument, which will ensure
that the module is capable of generating a key when no previous key
exists regardless of the value of `regenerate`. Previously, the task
would always run with the `partial_idempotence` value, and that obscured
a bug (#597) that would occur when it was set to `fail`. The bug was
fixed in the previous commit.
* Show timings with devel, and skip everything else.
* Move to other group.
* Try smaller SSH key size (i.e. make tests run faster).
* Add implicit size that now must be explicit.
* Change group of luks_device.
* Revert "Show timings with devel, and skip everything else."
This reverts commit 7b73f7e4d7.
* Update CI scripts to be more close to the ones in ansible-core.
* Extend CI matrix.
* Mark more VMs.
* Revert "Mark more VMs."
This reverts commit 8bc79af636.
* Disable alpine VMs for get_certificate due to httptester problems.
* Improve retrieval of cryptsetup version.
* ACME 'emulator' won't work on Alpine either.
* Improve luks test setup.
* Make sure wipefs is installed on Alpine.
* dmsetup (from device-mapper) is used by the tests.
* Fix bcrypt install failure handling.
* String, not float.
* openssl_privatekey_convert is not an action module.
* Update Python info.
* Try out which VMs can be used by now.
* Enable ACME tests on all VMs but Alpine; update comment.
* Adjust acme-tiny shebang.
* Remove new entries from CI matrix.
The issuer_uri is retrieved from the Authority Information Access field the same way as the OCSP responder URI is.
Handling is exactly the same since they reside in the same OID space and have the same data type.
Tests have also been added based on the integration test certificates.
Signed-off-by: benaryorg <binary@benary.org>
Signed-off-by: benaryorg <binary@benary.org>
* Extend tests to check for privatekey_content together with privatekey_passphrase.
* Also test privatekey_content for private keys without passphrases.
* ci: enable rhel9.0 tests for openssh_cert
* ci: allow openssh_cert second signature algorithm test for versions >8.7
* ci: narrowing condition to not attempt RSA1 signing exclusively on RHEL >=9
* ci: grouping and documenting condition
* Add RHEL 9.0 and FreeBSD 13.1 to CI.
* Add Ubuntu 22.04 and Fedora 36 to CI.
* Switch orders so that root doesn't have a SHA1 signature.
* Skip openssh_cert test on RHEL 9.0.
* Make it possible that pyOpenSSL isn't installed *at all*.
* Work with default.
* Prepare IDNA/Unicode conversion code. Use to normalize input.
* Use IDNA library first (IDNA2008) and Python's IDNA2003 implementation as a fallback.
* Make sure idna is installed.
* Add changelog fragment.
* 'punycode' → 'idna'.
* Add name_encoding options and tests.
* Avoid invalid character for IDNA2008.
* Linting.
* Forgot to upate value.
* Work around cryptography bug. Fix port handling for URIs.
* Forgot other place sensitive to cryptography bug.
* Forgot one. (Will likely still fail.)
* Decode IDNA in _compress_entry() to avoid comparison screw-ups.
* Work around Python 3.5 problem in Ansible 2.9's default test container.
* Update changelog fragment.
* Fix error, add tests.
* Python 2 compatibility.
* Update requirements.
* Fix empty check for openssl_pkcs12 tests.
* Remove unnecessary imports.
* Prevent crash if PyOpenSSL cannot be imported because of an AttributeError.
* Add changelog fragment.
* Fix constraints file.
* Use Python 2.7 instead of 3.5 for 2.9 cloud tests (pip module is broken).
* Prevent upgrading cryptography on ansible-core 2.12's default container with Python 3.9.
Read and write work queue significantly degrades performance on
SSD/NVME devices[1].
In Debian 11 crypttab does not support no-read-workqueue and
no-write-workqueue flags, so the persistent flag is workaround: once
opened with perf parameters persists forever.
[1] https://blog.cloudflare.com/speeding-up-linux-disk-encryption/
Signed-off-by: Yauhen Artsiukhou <jsirex@gmail.com>
* Use community ansible-test images.
* Adjust tests for new operating systems, and pass on Python version as well.
* Fix Python version.
Co-authored-by: David Moreau Simard <moi@dmsimard.com>
* Fix package name.
Co-authored-by: David Moreau Simard <moi@dmsimard.com>
* Allow multiple intermediate CAs to have same subject.
* Add tests.
* Fix test name.
* Don't use CN for SAN.
* Make a bit more compatible.
* Include jinja2 compat for CentOS 6.
Felix Fontein
David Zaslavsky
Katze
Andrew Pantuso
Andrew Pantuso
Yauhen