[BUG] Prospector CaptureStream triggers a false positive pylint check

[volans-]

Describe the bug

False positive pylint no-member reported due to Prospector overriding sys.stdout with a CaptureStream instance.

To Reproduce

1. Have some code that checks for sys.stdout.isatty()
2. Call prospector so that it runs pylint
3. Get as error: pylint: no-member / Instance of 'CaptureStream' has no 'isatty' member (col 15)

Expected behavior

No error is reported

Environment

• OS: macOS Sonoma
• Tool pylint
• Prospector version 1.12.0
• Python version 3.11 (reproduced also with 3.9)

[Eljees]

The reported symptom is gone, but the cause behind it is not, so it may be worth separating the two.

Symptom. I could not get no-member for sys.stdout.isatty() on prospector 1.19.1 (master 1814a2f) with pylint 4.0.6 / astroid 4.0.4 - 0 messages, and the same for encoding, fileno(), reconfigure() and buffer. isatty() was added to CaptureStream in 16af5ba while fixing #703, which covers exactly the member from the report.

Cause. While the tools run, astroid still resolves sys.stdout to the live object, not to typeshed:

with CaptureOutput(hide=True):
    astroid.extract_node("import sys\nsys.stdout #@").inferred()
# [Instance of prospector.tools.utils.CaptureStream]

So members are still checked against prospector's class rather than the real stream - they just happen to resolve now, through the TextIOWrapper base.

There is a second, non-static consequence of the same object: CaptureStream.__init__ never calls TextIOWrapper.__init__, so for the whole duration of a run sys.stdout.fileno(), .errors, .name and .readable() raise ValueError: I/O operation on uninitialized object and .encoding is None. That is the same failure mode as #703, and I have opened #837 for it.

Happy to dig further if anyone still has a reproducer for the original no-member.