| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |\
| |
| | |
serverspec 2 fixes
|
| | |
| |
| |
| | |
also some cleanup in the os-detection
|
| |/ |
|
| | |
|
| |\
| |
| | |
Match group func tests to specification
|
| | | |
|
| | |
| |
| |
| |
| |
| |
| | |
When a group does not exist, the group resource should raise an exception on action modify,
but not on action_manage.
Also update the `not_to raise_error(SpecificError)` deprecated syntax.
|
| |\ \
| | |
| | | |
Work around breaking change in git clone
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
In git 1.7.10 shallow clone behavior was changed from shallow cloning
all branches by default, to only shallow cloning the branch being
checked out. Unfortunately, this breaks prior and expected behavior in
Chef as, on subsequent deploys it is not possible to switch branch,
because no other branches exist in the cache-copy of the repo.
To fix this we must use the `--no-single-branch` flag when performing
the inital clone. Unfortunately this flag is not supported by git
versions prior to 1.7.10, so we must check the version of git before
deciding if the workaround is required.
It probably makes sense to make this prior behavior optional in a future
patch, however this one just seeks to give consistent results across all
versions of git.
|
| |\ \ \
| |_|/
|/| | |
add changelog for 50x errors
|
| |/ / |
|
| |\ \
| | |
| | | |
retry on HTTP 50X Error when calling Chef REST API
|
| | | |
| | |
| | |
| | | |
the log.
|
| | | | |
|
| |\ \ \
| |_|/
|/| | |
Deprecate CookbookVersion#latest_cookbooks
|
| | | | |
|
| | | | |
|
| |\ \ \
| | | |
| | | | |
Disable unforked interval runs.
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
| | | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Exit gracefully when sent TERM
Fix spec.
|
| | | | | |
|
| |/ / /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Clarify error message.
Move fork and interval logic to application (specs pending).
Clean code logic, fix specs.
Allow unforked client runs with daemonize or splay
|
| | | | |
|
| | | |
| | |
| | |
| | |
| | | |
We now raise a little late when given a guard interpreter and a block, mostly
being lazy. This updates the functional test to check the correct code path.
|
| | | |
| | |
| | |
| | |
| | | |
These tests are no longer correct since we don't allow passing a block with
a guard interpreter.
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | | |
The fix for #1943 moved some code into a new configure method, which is also
where we now raise if we're passed a block and a guard_interpreter as that is
where we evaluate guard interpreters. This updates the unit test for that
behavior.
|
| | | |
| | |
| | |
| | |
| | | |
We already check for command || block_given? in initialize, so
command cannot be nil && not have a block when we get here.
|
| | | |
| | |
| | |
| | |
| | | |
Refactor the unit tests to not test Chef::GuardInterpreter::ResourceGuardInterpreter
directly rather than through Chef::Resource.
|
| | | |
| | |
| | |
| | |
| | |
| | | |
make sure we don't evaluate the guard_interpreter immediately, i.e. during
compilation of the resource, as the conditional may be seen before the
guard_interpreter attribute.
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
If we don't create the guard_interpreter until we're ready to test it,
we're sure to already have all the resource attributes evaluated.
Previously we set up the guard_interpreter upon initialization, that
is when it was first set on the resource.
|
| | | |
| | |
| | |
| | |
| | | |
adds a functional test to ensure guard_interpreter can be set after a
conditional.
|
| | | |
| | |
| | |
| | |
| | |
| | | |
reorder the powershell functional tests to provide a framework for testing
with guard_interpreter set and not set (allowing testing the "default",
i.e. values on resource initialization).
|
| | | |
| | |
| | |
| | |
| | | |
The tests that ensured we didn't run a conditional when it wasn't specified
had no expectations, so they would always have passed.
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | | |
we now raise an error when passed a block and a guard_interpreter is
specified other than :default. When other interpreters become the "default"
(which actually means use the default shell for that platform) we will
need to consider those as well, i.e. #1495.
|
| |/ /
| |
| |
| |
| |
| |
| |
| | |
Raises an exception if guard_interpreter is set (not :default) and we are not
given a command (i.e. we are given a block). This is not supported, as we pass
a command to an external interpreter.
Related to #1943
|
| |\ \
| | |
| | | |
Jdmundrawala/provider specs
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| |\ \ \
| | | |
| | | | |
Make knife unit tests pass on windows
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
| | | | | |
|
