Fix nested subscript assignment inference

[AG0708]

Type of Changes

	Type
✓	🐛 Bug fix

Description

Pylint could infer the original None value from a nested dictionary literal even when the same key was assigned a dictionary immediately before the nested assignment. That produced unsupported-assignment-operation for valid code.

This adds a narrow fallback for adjacent assignments to the exact same literal-key chain when the root is inferred as a built-in dictionary or list. Custom mappings and non-subscriptable replacement values still emit E1137; the regression suite covers both cases.

Validation:

• functional suite: 892 passed, 12 skipped
• checker suite: 406 passed, 44 skipped, 4 xfailed
• full pre-commit run --all-files: passed
• towncrier draft: passed

Closes #10050

Implementation and tests were developed with OpenAI Codex assistance and reviewed against custom-mapping and aliasing edge cases before submission.

[Pierre-Sassoulas]

Thank you for contributing to pylint. Not looking too much into the details, it seems it adds a lot of complexity. Could it be 'hidden' by a better astroid inference ?

[AG0708]

Thanks—that makes sense. I confirmed the underlying behavior is Astroid inference: at the final assignment, test_dict["key"] still infers to the original None and does not see the immediately preceding subscript mutation. I’m investigating whether this can be expressed as a smaller Astroid-level inference improvement (or a reusable inference helper) so the checker does not carry the mutation logic itself. I’ll update the patch after checking the soundness boundaries for custom mappings and properties.

[AG0708]

Follow-up: I opened pylint-dev/astroid#3220 with the narrow inference fix, including downstream Pylint regression coverage and adversarial custom-mapping/property cases. If that lands, I’ll simplify this PR to remove the checker-side inference fallback and retain only the Pylint regression/news plus the required Astroid dependency update.

[Pierre-Sassoulas]

Thank you for working on pylint, I'm going to add some test cases, it's going to be easier than describing them :)

[codecov]

Codecov Report

❌ Patch coverage is 96.29630% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 96.40%. Comparing base (5d7d0cb) to head (dc9e001).
⚠️ Report is 12 commits behind head on main.

Files with missing lines	Patch %	Lines
pylint/checkers/typecheck.py	96.29%	1 Missing ⚠️

❌ Your patch check has failed because the patch coverage (96.29%) is below the target coverage (100.00%). You can increase the patch coverage or adjust the target coverage.

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #11255      +/-   ##
==========================================
+ Coverage   96.36%   96.40%   +0.04%     
==========================================
  Files         178      178              
  Lines       19952    20069     +117     
==========================================
+ Hits        19227    19348     +121     
+ Misses        725      721       -4     

Files with missing lines	Coverage Δ
pylint/checkers/typecheck.py	96.57% <96.29%> (-0.01%)	⬇️

... and 7 files with indirect coverage changes

🚀 New features to boost your workflow:

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

[github-actions]

🤖 According to the primer, this change has no effect on the checked open source code. 🤖🎉

This comment was generated for commit dc9e001

[AG0708]

Thanks—the added cases made the limitation clear. I generalized pylint-dev/astroid#3220 in 6546320 so the inference layer now handles the six new shapes here: intervening statements, reads, deletes, deeper chains, aliases, and ambiguous writes.

Running this branch against that Astroid commit removes all six newly documented false positives. Astroid's 483 inference tests and all changed-file hooks pass; its remaining full-suite failures reproduce unchanged on the parent commit.

Once the Astroid change is available, I will remove the checker-side fallback from this PR and keep the Pylint functional coverage and release note.

[Pierre-Sassoulas]

Could add some coverage or remove the code that can't be reached, please ?