Do not crash when an error message contains format placeholders

[Sanjays2402]

Type of Changes

	Type
✓	🐛 Bug fix

Description

AstroidError.__str__ runs str.format on its message, and several inference errors build that message out of user code. namedtuple("{0}", "") therefore reached "ValueError: " + str(exc) with an unsatisfiable positional placeholder and raised IndexError instead of the intended UseInferenceDefault, crashing pylint.

The fallback already caught ValueError; it now also catches IndexError and KeyError, the other two exceptions str.format raises for placeholders it cannot fill. A regression test was added beside the sibling test_*_does_not_crash_inference cases in tests/brain/test_named_tuple.py (fails without the fix with the reported IndexError, passes with it).

Happy to add a ChangeLog entry if you would like one for this.

This change was prepared with AI assistance; the regression test was run locally and fails without the fix.

Closes #3199

[codspeed-hq]

Merging this PR will not alter performance

✅ 3 untouched benchmarks
⏩ 1 skipped benchmark¹

---

_{Comparing Sanjays2402:fix/error-str-format-placeholders (1264770) with main (8666418)}

Footnotes

1. 1 benchmark was skipped, so the baseline result was used instead. If it was deleted from the codebase, click here and archive it to remove it from the performance reports. ↩

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 93.64%. Comparing base (8666418) to head (1264770).
⚠️ Report is 8 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #3203   +/-   ##
=======================================
  Coverage   93.64%   93.64%           
=======================================
  Files          93       93           
  Lines       11567    11567           
=======================================
  Hits        10832    10832           
  Misses        735      735           

Flag	Coverage Δ
linux	93.50% <100.00%> (ø)
pypy	93.64% <100.00%> (ø)
windows	93.61% <100.00%> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
astroid/exceptions.py	95.83% <100.00%> (ø)

🚀 New features to boost your workflow:

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

[DanielNoord]

Can we cover the KeyError with a test as well?

[Pierre-Sassoulas]

Two payloads still get through:

• namedtuple("{message.foo}", "") => AttributeError: 'str' object has no attribute 'foo'
• namedtuple("{message:200000000}", "") => no exception at all; format succeeds and builds a 200 MB string. Peak RSS goes from 22 MB to 421 MB for a single call, so no except clause can catch this one.

#3214 covers both by also dropping message from the substitution fields, which stops the message from being able to interpolate itself (Please do it that way and credit @kali834x for it).

Also adding the changeLog entry would be welcome, plus the KeyError test Daniel asked for.

[Sanjays2402]

Implemented the requested hardening in 889342c: AstroidError now excludes the message itself from format fields and falls back on LookupError/ValueError/AttributeError/TypeError, which prevents both attribute traversal and large-width self-formatting. Added the KeyError and AttributeError inference regressions plus a ChangeLog entry. The full tests/brain/test_named_tuple.py passes (26 passed); with the source fix reverted, the AttributeError regression fails during AstroidError.str, and it passes again with the fix.

[DanielNoord]

Please fix CI and rebase.