This fixes a regression I introduced the other day in #40481 when I
backported auth fixes to 2016.3. I changed how the client was
instantiated and ended up passing an unsupported kwarg to the wrapped
function. This resolves that regression.
os.makedirs() will raise OSError in case the directory passed as argument
already exists. We do check that this is not the case right before the
call, but there is still a tiny time window in which the directory might
be concurrently created between the isdir() check and the makedirs() call.
In some unlucky cases under heavy I/O the following stack trace is produced:
The minion function caused an exception: Traceback (most recent call last):
...
File "/usr/lib/python2.7/site-packages/salt/fileclient.py", line 165, in cache_file
return self.get_url(path, '', True, saltenv, cachedir=cachedir)
...
File "/usr/lib/python2.7/site-packages/salt/fileclient.py", line 126, in _cache_loc
os.makedirs(destdir)
File "/usr/lib64/python2.7/os.py", line 157, in makedirs
mkdir(name, mode)
OSError: [Errno 17] File exists: <PATH>
This will force a re-download if the hash specified by source_hash if it
doesn't match the hash of the cached file, but if the hash matches, it
will skip the download. This will permit keep=True to prevent repeated
downloads of the source file, if and only if the source_hash also
remains the same.
First of all, the opts will never not have this key, as it has a default
value in salt/config/__init__.py. Second, the default hash type changed
to sha256 anyway in 2016.11, so we shouldn't be referring to a specific
hash type when it'll no longer be accurate after a merge-forward.
1. The trim_output argument was ignored for archives extracted using
tarfile.
2. file.get_source_sum would fail if "source" is a list
3. The checksum was being updated before we checked to see if it
matched, effectively keeping us from detecting changes to the hash.
4. When source_hash_update is True, and the archive is extracted, and
files are later removed from the extraction dir, the state would not
re-extract the archive unless the source_hash had changed.
Fixes#40005
This new option allows the user to specify what kind of hash was
used when the public key was hashed to a fingerprint. The default
hash is md5, but the user should be able to specify this if they
know what the fingerprint was originally hashed with.
This prevents the ssh_minion_opts from overwriting values that should
have been inherited or otherwise setup by the master for the SSH minion.
Resolves#39892
The fact that PyYAML was the last renderer in the render pipe initially,
masked the fact that template renderers return StringIO instances and
not strings. This commit makes sure that we try reading from the
StringIO first and then fallback to just trying to base64 encode the
userdata value as returned by compile_template().
