gh-85287: Convert UnicodeError to UnicodeEncodeError| UnicodeDecodeError

[srinivasreddy]

https://bugs.python.org/issue41115

• Issue: Codecs should raise precise UnicodeDecodeError or UnicodeEncodeError #85287

[eamanu]

good catch

[eamanu]

LGTM

[srinivasreddy]

@pitrou Please review!

[doerwalter]

The patch get the offsets in the exception object wrong for all non-trivial cases.

For example the following case:

>>> s = 'foo.' + 60*'\xff'
>>> s.encode('idna')

gives the stack trace:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/Users/walter/checkouts/cpython-doerwalter/Lib/encodings/idna.py", line 187, in encode
    result.extend(ToASCII(label))
  File "/Users/walter/checkouts/cpython-doerwalter/Lib/encodings/idna.py", line 105, in ToASCII
    raise UnicodeEncodeError("punycode", label.decode("punycode"), 0,
UnicodeEncodeError: 'punycode' codec can't encode characters in position 0-62: label too long

However the bad characters are not s[0:63] but s[4:64].

Fixing this would require to track the correct offset across multiple function calls. I'm not sure whether that added complexity is justified. idna enforces strict encoding anyway.

[arhadthedev]

A review by @doerwalter hasn't been being addressed for three years and the OP blocked their branch from external pushes thus preventing us from applying all necessary fixes by ourselves.