tarfile silently stops/skips on bad member header (e.g. checksum mismatch)

Bug report

Bug description:

When reading a tar archive that includes a file with a bad header (such as a checksum mismatch), getmembers simply stops listing the members at that file, without reporting an error, and ignoring the files that come after it (Edit: unless ignore_zeros=True is set).

I would expect instead that getmembers lists all members, and extractfile raises a TarError when trying to extract a file with an invalid header (such as a bad chksum or typeflag).

import os
import contextlib
import subprocess
from io import BytesIO
from tarfile import TarFile, TarInfo
from tempfile import TemporaryDirectory

# generate a tar file in memory
bio = BytesIO()
with TarFile(mode="w", fileobj=bio, errorlevel=2) as tf:
    ti = TarInfo()
    ti.size = 3
    for name, data in (("foo", b"123"), ("bar", b"456"), ("quz", b"789")):
        ti.name = name
        tf.addfile(ti, BytesIO(data))

# break the checksum of the second file 'bar'
assert b"\x00006425\x00" in bio.getvalue()
broken = bio.getvalue().replace(b"\x00006425\x00", b"\x00106425\x00")

# try to read the tar file
with TarFile(fileobj=BytesIO(broken), errorlevel=2) as tf:
    for ti in tf.getmembers():
        print(repr(ti.name))
        with tf.extractfile(ti) as fh:
            print(repr(fh.read()))
# => only "foo" is extracted

with TemporaryDirectory() as td:
    with contextlib.chdir(td):
        with TarFile(fileobj=BytesIO(broken), errorlevel=2) as tf:
            tf.extractall()
        print(os.listdir())
        # => again only "foo" is extracted
        os.unlink("foo")
        with TarFile(fileobj=BytesIO(broken), errorlevel=2) as tf:
            tf.extractall(filter="data")
        print(os.listdir())
        # => filter doesn't change anything
        with open("test.tar", "wb") as fh:
            fh.write(broken)
        subprocess.run(["tar", "tvf", "test.tar"], check=False)
        # => GNU tar 1.34 correctly identifies error and continues processing

Output:

'foo'
b'123'
['foo']
['foo']
-rw-r--r-- 0/0               3 1970-01-01 00:00 foo
tar: Skipping to next header
-rw-r--r-- 0/0               3 1970-01-01 00:00 quz
tar: Exiting with failure status due to previous errors

CPython versions tested on:

3.12, 3.13

Operating systems tested on:

Linux, Windows

Reproduisez le comportement avec l’exemple BytesIO fourni, puis partez de TarFile.getmembers(), extractfile() et extractall(), les points d’entrée nommés dans le rapport. Suivez la manière dont un en-tête de membre invalide est traité et dont les en-têtes suivants sont atteints. Le travail est considéré comme terminé lorsque le mauvais en-tête est signalé comme demandé, que les membres valides suivants restent traitables et que le comportement concerné est couvert par des tests.