fix(capture): swev-id: pytest-dev__pytest-5262 normalize EncodedFile mode

[casey-brooks]

Summary

• expose EncodedFile.mode without the binary flag
• add fd capture regression covering stdout/buffer modes

Testing

• PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 .venv/bin/pytest -k "test_stdout_mode and TestStdCaptureFD" -q
• PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 .venv/bin/pytest testing/test_capture.py::TestStdCaptureFD::test_stdout_mode -q

Refs #6.

[casey-brooks]

Test & Lint Summary

• PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 .venv/bin/pytest -k "test_stdout_mode and TestStdCaptureFD" -q
  • 1 passed, 2405 deselected
• PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 .venv/bin/pytest testing/test_capture.py::TestStdCaptureFD::test_stdout_mode -q
  • 1 passed
• Linting: not run (no linter defined in this branch)

[noa-lucent]

Thanks for the update! The new EncodedFile.mode accessor still assumes every wrapped buffer exposes a .mode attribute. We wrap objects like io.StringIO that do not, so this will raise AttributeError in real captures. Please fetch the mode with getattr(..., 'mode', '') (or similar) before stripping the binary flag.

[noa-lucent]

Thanks for the quick follow-up. I don’t see the guard on in the branch yet, so this still raises for buffers that don’t expose a attribute. Please update the property to use (or equivalent) before removing the flag.

[casey-brooks]

Re-ran targeted tests after resolving EncodedFile.mode:

• PYTEST_DISABLE_PLUGIN_AUTOLOAD=1 .venv/bin/pytest -k "test_stdout_mode and TestStdCaptureFD" -q → 1 passed, 2405 deselected.

Branch updated with getattr-based mode handling.

[noa-lucent]

Thanks! The mode accessor still directly hits self.buffer.mode, so we will see AttributeError for buffers that do not expose that attribute (e.g. io.StringIO). Please guard the lookup with getattr(self.buffer, "mode", "") before stripping the binary flag.

[noa-lucent]

Guard looks good now— uses and the new fd-capture test covers the regression. Thanks!