| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
| |
We need to check if the env variable is set to empty string. If we don't we can get in an edge case where we blow up trying to call URI.parse.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
See:
* https://www.kickflop.net/blog/2013/06/27/syslog-output-for-chef-runs/
* http://lists.opscode.com/sympa/arc/chef/2014-12/msg00152.html
* https://github.com/higanworks-cookbooks/chef-client_syslog
It is relatively embarrassing to have a system administration tool that
doesn't support syslog out of the box.
|
| |
|
|\
| |
| | |
Apply an SSL Policy to CookbookSiteStreamingUploader, fixing SSL errors uploading to private Supermarkets
|
| | |
|
| |
| |
| |
| | |
goodness
|
| |
| |
| |
| | |
for us
|
| |
| |
| |
| | |
uploading to private Supermarkets
|
|\ \
| | |
| | | |
Convert wiki links to docs.chef.io links
|
| | | |
|
| | | |
|
|\ \ \ |
|
| | | | |
|
| |\ \ \
| | |_|/
| |/| |
| | | | |
into minshallj-minshallj/keep_suid_bit
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
On Linux updating the group or owner unsets the suid bit for security
reasons, so check for group and owner updates whether or not to set the
mode.
|
|/ / /
| | |
| | |
| | |
| | | |
Closes #3327
Closes #3248
|
| | | |
|
| | | |
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This takes advantage of Chef::HTTP::Authenticator.handle_request. After the authentication_headers have been merged, it adds:
{'X-Ops-Server-API-Version' => Chef::HTTP::Authenticator::SERVER_API_VERSION}
Chef::HTTP::Authenticator::SERVER_API_VERSION should corrospond to the max version of the server the client is compatible with. We are starting at "0".
I also removed outer level lets :base_header and :req_with_body_headers from rest_spec.rb as they were never used anywhere. Another :base_header was defined in a tighter part of the spec and was only used in that scope, and :req_with_body_headers was never used anywhere.
|
|\ \ \
| | | |
| | | | |
Lcg/node utf8 sanitize
|
| | | | |
|
| | |/
| |/|
| | |
| | |
| | |
| | |
| | | |
with validation off any bad utf8 data will not cause the
ffi-yajl encoder to raise. with ffi-yajl >= 2.2.0 the bad
data will be scrubbed to produce valid JSON to POST/PUT to
the server.
|
|\ \ \
| |/ /
|/| | |
fix AIX package installs using a 'source' attribute
|
| | |
| | |
| | |
| | | |
that package
|
| | | |
|
| | |
| | |
| | |
| | | |
Fixes #2778
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
<< has higher precedence than == & ?:
=> "mdbtools >= 0.7.1: Unable to match package 'mdbtools >= 0.7.1' but matched 1 \u0001"
Obvious fix.
|
| | | |
|
| | |
| | |
| | |
| | | |
and fix it to hit the right key.
|
| | |
| | |
| | |
| | |
| | |
| | | |
Use this to override the state reported by the resource reporter
while avoiding the collision over Chef::Resource#state being used
by some LWRPs.
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
this line violates the rule that we never mutate the new resource.
in Chef 12 this causes real problems because if a file resource is
notified (run twice) then the rendered content is loaded into the
new resource and if the file resource content changes (i.e. its a
template with logic which changes when its notified) then it will
cause a failure because the rendered content will not match the
'requested' checksum. where the 'requested' checksum is actually
the checksum loaded by this line.
fundamentally we have three different states that we're trying to
track:
- current state
- initial state
- requested state
the removed line tries to use the @new_resource for reporting initial
state, which then bleeds over into what chef thinks is requested state
in the next invokation.
without constructing a third @initial_resource and using that for
resource reporting comparison we can't solve this problem "right".
the tradeoff is we either break reporting here or break chef-client
runs. this patch sacrifices reporting in order to make chef-client
work.
|
|\ \ \
| | | |
| | | | |
Show Chef::VERSION at prompt_c and prompt_i on shell session
|
| | | | |
|
| |\ \ \ |
|
| | | | | |
|
|\ \ \ \ \
| | | | | |
| | | | | | |
Implemented `knife user key edit` and `knife client key edit`
|
| | | | | | |
|
| | | | | | |
|
| | | | | | |
|
| | | | | |
| | | | | |
| | | | | |
| | | | | | |
from server on update.
|
| | | | | | |
|
|/ / / / /
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | |
| | | | | |
This was broken in #3034. The issue is that the package does not get detected
as installed because the version that is installed has \x00 at the end,
while the version from the msi does not. This fails comparison, and we fall
into code that does not use source and thus requires candidate version and we
die. (Or something like that)
Solves #3316
|