Basically, we were taking the retcode from opts_pkg calls and then
returning it as the retcode for the wfunc call. This is misleading as
opts_pkg is usually successful. I also think it was the only piece of
code utilizing the line I removed in 439a2799ce
The FunctionWrapper class is primarily used for cross-calling modules
from wfuncs and templating. Inserting the retcode there is probably not
helpful (it likely won't bubble through, because most calling contexts
won't look for it), and plus, we're assuming the return type is a dict
here, which is often not the case. I think this line was added
erroneously.
@ngrennan-inflection Could you look at this please? Curious about your
thoughts, and whether you tested that this specific line was actually
helping your new retcode-handling code.
The `ret` datastructure sometimes looked like this:
{'basepi02': "Target 'basepi02' did not return any data, probably due to an error."}
Return information is not a dict, handle that case, assume it's an error
Per https://github.com/saltstack/salt/issues/23567 there is an exception
when trying to parse headers. The values in the returned ret['headers']
now defaults to just being a list of strings containing the keys, but no
values are included. It seems that in the past a list of strings of
key:value pairs was returned.
This patch requests full_headers, which instead returns a dictionary of
header keys and values. The keys are lowercased, but otherwise they
seem to be the same.
So this patch requests the full_headers and deals with lowercasing the
desired headers. This works for me as long as both
8999148da0
and
80c7a697ae
are applied.
since salt.utils.sdecode() is bias toward system encoding(defaults to cp1252) and it differs from the strings returned from _get_reg_software() (utf-8)
(#25532)
This implements the solution to #25477. Any ``include`` directives
within state files can use multiple dot notation (influenced by Python
relative imports) to refer to a parent state.
This also implements the same functionality inside the stateconf
renderer
(http://docs.saltstack.com/en/latest/ref/renderers/all/salt.renderers.stateconf.html),
which supported relative includes, excludes, and prefixed states. This
PR expands on existing tests for stateconf.
This also updates the documentation accordingly.
Per FHS, /var/tmp contains temporary files and directories that are not
cleaned up across system reboots. It looks like some commands in the
bootstrap process resulted in "boots" of the LXC inadvertently cleaning
/tmp while its content was still needed. No such problem occurs when
/var/tmp is used.
