PG-1504 Make partitions inherit encryption status

[AndersAstrand]

Since the access method isn't included in the statement we'll have to look it up on the parent in the same way DefineRelation() does.

[codecov-commenter]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 78.45%. Comparing base (bb000ef) to head (050d87e).
Report is 1 commits behind head on TDE_REL_17_STABLE.

❌ Your project status has failed because the head coverage (78.45%) is below the target coverage (90.00%). You can increase the head coverage or adjust the target coverage.

Additional details and impacted files

@@                  Coverage Diff                  @@
##           TDE_REL_17_STABLE     #253      +/-   ##
=====================================================
+ Coverage              78.32%   78.45%   +0.13%     
=====================================================
  Files                     22       22              
  Lines                   2468     2479      +11     
  Branches                 385      387       +2     
=====================================================
+ Hits                    1933     1945      +12     
  Misses                   459      459              
+ Partials                  76       75       -1     

Components	Coverage Δ
access	79.35% <ø> (ø)
catalog	86.32% <ø> (ø)
common	92.50% <ø> (ø)
encryption	71.90% <ø> (ø)
keyring	72.07% <ø> (ø)
src	55.73% <100.00%> (+2.09%)	⬆️
smgr	98.01% <ø> (-0.02%)	⬇️
transam	∅ <ø> (∅)

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[jeltz]

I think it has a bug, but even if not I think the tests should also involve the GUC.

[dutow]

I'm not 100% sure how to handle enforce_encryption with this.
Currently by looking at the code, if it is set, we can't create new partitions for not encrypted tables.
Which probably makes sense, but on the other hand we do allow to do other modifications of non encrypted tables, like creating new indexes.

At least the testcase could include this behavior.

[AndersAstrand]

I'm not 100% sure how to handle enforce_encryption with this. Currently by looking at the code, if it is set, we can't create new partitions for not encrypted tables. Which probably makes sense, but on the other hand we do allow to do other modifications of non encrypted tables, like creating new indexes.

At least the testcase could include this behavior.

Agreed! I included the current behavior in the test, we can change it later if we would want to allow creation of unencrypted partitions even if the GUC is set.

[jeltz]

I think there is one case with the default GUC which is still not properly handled but I could be wrong.

[AndersAstrand]

I think there is one case with the default GUC which is still not properly handled but I could be wrong.

Should be fixed now!

[jeltz]

You could accomplish the same with relation_open_rv() but I am not sure which is cleaner.

[jeltz]

I think PostgreSQL style for this is foundAccessMethod = parentAmOid != InvalidOid.

[AndersAstrand]

Yes! That's better. I didn't know about InvalidOid so i felt this was just as good as randomly comparing to 0

Pushed a fixup and also added a noop to the if just below to make it slightly clearer what's going on.