If an object ob does not have a __bytes__ method or buffer interface and is not a string, integer, or iterable, bytes(ob) fails with 

TypeError: 'object' object is not iterable

This is misleadingly narror (similar to #17032). We should either list *all* the things that ob is not

'object' object is not a string, integer, or iterable and does not have a __bytes__ method or buffer interface

or simply say

'object' object cannot be converted to bytes

and let the programmer recheck what can be.

See also #16518.

Do you prefer the long or short form?

Short form for me.

> 'object' object cannot be converted to bytes

This is not entirely accurate for the bytes(int) case though, since it's not converting the int to bytes.  OTOH if we want to keep the message short we have to make compromises about accuracy and completeness.

Here is a patch.

I think this patch is okay.

I would have suggested “Cannot construct bytes from [. . .]” to avoid the problem with “convert”, but this message is produced in places other than the constructor, e.g.

>>> int.from_bytes("str", "little")
TypeError: cannot convert unicode object to bytes

Thank you Martin.

Interesting, int.from_bytes() accepts lists (and other iterables):

>>> int.from_bytes([1, 2, 3], "little")
197121

Is it intentional?

New changeset 6b16eec56854 by Serhiy Storchaka in branch 'default':
Issue #17339: Improved TypeError message in bytes constructor.
https://hg.python.org/cpython/rev/6b16eec56854

The int.from_bytes() behaviour seems to have been documented from the beginning, so maybe it was intended.