[Hisernberg]

Description

BuddyNSResponse, DNSPodResponse, LuadnsResponse and NsOneResponse declare errors/objects as class-level mutable attributes and append to them from parse_body_and_errors(). The lists are therefore shared by every response object in the process:

• parsed objects accumulate across requests: libcloud/dns/drivers/nsone.py get_zone() reads response.objects[0] and can receive data from an earlier response;
• errors leak between requests, so success() can report stale failures.

Reproduced locally against trunk:

from types import SimpleNamespace
from libcloud.common.buddyns import BuddyNSResponse

def raw(body):
    return SimpleNamespace(headers={}, body=body, text=body, status_code=200,
                           reason="OK", request=None, iter_content=lambda *a: iter([]))

r1 = BuddyNSResponse(response=raw('{"zones": [1]}'), connection=None)
r2 = BuddyNSResponse(response=raw('{"zones": [2]}'), connection=None)
print(r2.objects)   # [{'zones': [1]}, {'zones': [2]}]  <- stale first response inside r2

Fix

Initialize fresh per-instance errors/objects lists before super().__init__() (the base Response constructor calls success(), which reads errors), and convert the class attributes to bare annotations. Per-request behavior is otherwise unchanged.

Tests

Regression tests added to all four driver test suites (libcloud/test/dns/test_buddyns.py, test_dnspod.py, test_luadns.py, test_nsone.py):

• test_objects_do_not_accumulate_across_responses
• test_errors_do_not_leak_across_responses

With the old code the four suites report 4 failures + 1 error; with this change all 60 tests pass.

AI Generation Disclosure

Developed with AI assistance (agent workflow): the bug was found by an AST scan for mutated class-level attributes, reproduced against trunk, fixed, and verified with the A/B test run above.

[micafer]

Thanks for the fix.

There is some overlap with PR #2066 in NsOneResponse, but I don't consider that a blocker for this focused fix. We can reconcile the overlapping change when updating #2066 against trunk.

One minor, suggestion: the new test classes are currently placed after the if __name__ == "__main__": block. Moving them above that block would ensure they are also discovered when the test modules are executed directly.

Please also add the entry in the changelog.

[codecov-commenter]

Codecov Report

❌ Patch coverage is 95.00000% with 4 lines in your changes missing coverage. Please review.
✅ Project coverage is 83.65%. Comparing base (d3bb996) to head (13faa0e).
⚠️ Report is 13 commits behind head on trunk.

Files with missing lines	Patch %	Lines
libcloud/test/dns/test_buddyns.py	93.75%	0 Missing and 1 partial ⚠️
libcloud/test/dns/test_dnspod.py	93.75%	0 Missing and 1 partial ⚠️
libcloud/test/dns/test_luadns.py	93.75%	0 Missing and 1 partial ⚠️
libcloud/test/dns/test_nsone.py	93.75%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##            trunk    #2216      +/-   ##
==========================================
+ Coverage   83.64%   83.65%   +0.01%     
==========================================
  Files         352      352              
  Lines       82321    82393      +72     
  Branches     8868     8880      +12     
==========================================
+ Hits        68855    68921      +66     
- Misses      10593    10594       +1     
- Partials     2873     2878       +5     

Files with missing lines	Coverage Δ
libcloud/common/buddyns.py	100.00% <100.00%> (ø)
libcloud/common/dnspod.py	100.00% <100.00%> (ø)
libcloud/common/luadns.py	100.00% <100.00%> (ø)
libcloud/common/nsone.py	100.00% <100.00%> (ø)
libcloud/test/dns/test_buddyns.py	96.23% <93.75%> (-0.44%)	⬇️
libcloud/test/dns/test_dnspod.py	96.65% <93.75%> (-0.28%)	⬇️
libcloud/test/dns/test_luadns.py	96.00% <93.75%> (-0.20%)	⬇️
libcloud/test/dns/test_nsone.py	94.42% <93.75%> (-0.06%)	⬇️

... and 1 file with indirect coverage changes

🚀 New features to boost your workflow:

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

[Hisernberg]

Done, thanks @micafer:

• the four regression test classes now sit above the if __name__ == "__main__": guard (verified they run when the modules are executed directly)
• changelog entry added under the 4.0.0 DNS section
• all 64 tests in the four files pass locally

Re: the NsOne overlap with #2066 — happy to reconcile once that one is updated against trunk.

[Hisernberg]

Reconciliation analysis for the #2066 overlap (@aaron-sierra cc @micafer):

The NsOneResponse hunks in #2066 and here are semantically identical — both move errors/objects from class attributes to per-instance lists assigned before super().__init__() (which matters: JsonResponse.__init__ calls success(), which reads self.errors), then reassign them from parse_body_and_errors(). This PR additionally annotates the attributes, documents the ordering constraint, applies the same fix to BuddyNS/DNSPod/LuaDNS, and pins the behavior with regression tests (test_errors_do_not_leak_across_responses, test_objects_do_not_accumulate_across_responses in libcloud/test/dns/test_nsone.py).

So whichever PR merges first, the other rebases with at most a same-content conflict — no behavioral decision is needed. If #2066 lands first, its NsOneResponse hunk can simply be dropped in favor of this one (the regression tests here still apply unchanged).