gh-102799: Let pydoc use the exception instead of sys.exc_info

[iritkatriel]

• Issue: some usages of sys.exc_info can be replaced by more modern code #102799

[terryjreedy]

Except for suggestions, looks good. I satisfied my question.

[terryjreedy]

Separate suggestions:

1. Keep handling of two args in order. First first and second second. Handling first in middle of second is unnecessary confusion.
2. Handle two cases of second separately. Don't create tuple just to unpack. Unpacking makes it slightly more obvious that same attributes are defined, but pack/unpack makes it harder to see what object is assigned to each attribute.

Suggested change

	if not isinstance(exc_info, tuple):
	assert isinstance(exc_info, BaseException)
	exc_info = type(exc_info), exc_info, exc_info.__traceback__
	self.filename = filename
	self.exc, self.value, self.tb = exc_info
	self.filename = filename
	if isinstance(exc_info, tuple):
	self.exc, self.value, self.tb = exc_info
	else:
	assert isinstance(exc_info, BaseException)
	self.exc = type(exc_info)
	self.value = exc_info
	self.tb = exc_info.__traceback__

I am tempted to think that no user will ever raise this exception. pydoc is only documented as a command-line app, not as importable library. But this is not idlelib. Can/should we emit a DeprecationWarning in the isinstance clause? It could then be removed someday. Since there is no doc of the module content, no changed notice is needed.

[terryjreedy]

I personally prefer 'err' to 'e'. I have no idea of any general consensus.