Fix #38558 - pillar.get with default= ...,merge=true influence subsequent calls of pillar.get

[zwo-bot]

What does this PR do?

Fixes the unexpected behavior when calling pillar.get with merge=true described in #38558 by creating a copy of given default before calling salt.utils.dictupdate.update(default, ..)

What issues does this PR fix or reference?

#38558

Previous Behavior

If a pillar.get with a default and merge=true was called, subsequent calls of pillar.get will be return wrong results (depending on the previous merge)

Tests written?

No

[cachedout]

@terminalmage I believe you've been involved in this area recently?

[terminalmage]

This looks good. I can see another unrelated issue, but I'll open a separate PR to take care of it.

[terminalmage]

@rallytime This needs a backport to 2016.3.

[terminalmage]

#38601 fixes the unrelated issue I mentioned above, where using merge=True with a non-dict default would cause an error.

[multani]

I added some more commits containing unit tests for this bug in https://github.com/multani/salt.git , branch fix-issue-#38558:

• multani@a5c47b8
• multani@22f541f

That would be cool if the could be merged as part of this PR as well.

[terminalmage]

@multani You should be able to submit a pull request against @zwo-bot's fork.

[multani]

@terminalmage Oh yes, right, I didn't thought about that! It's done now!

[zwo-bot]

@multani Thx for writing the tests. You are right about the position of the deepcopy.

[cachedout]

Very nice! When the tests finish, I will get this in.

[frleb]

When will you update packages in your apt repository ? What does it trigger the update for the os packages ?

I tried to compare the default_settings and formula_settings, and the default_settings got identical values to the formula_settings.

{% set formula_settings = salt['pillar.get'](
'formula',
default=default_settings.formula,
merge=True
)
%}

I tried your master branch, and it fixes the problem.

Thanks.

FL