The documentation of readframes (wave module) states that readframes returns a string of bytes, just like python 2.7, but that's not the case

in python 3.4 readframes returns byte type objects

Here is the patch.

This is not the only place where "string of bytes" or "byte string" is used as a synonym to "bytes object".

Yes, and I think it would be good to consistently use "bytes object" or similar wording, to avoid confusion.  But not high enough priority to do a hunt-and-fix (unless someone really feels like doing that).  I've got a patch pending for fixing it in the base64 module, as part of making other fixes to that doc, for example.

Here is the patch replacing exact match 'string of bytes' with :class:`bytes` where appropriate. No match was found in base64.

Two of the changes don’t seem right to me; see the review. In the other two cases, IMO “string of bytes” is a reasonable description of :class:`bytes`, but being explicit is better.

Agree with Martin.  Only the two audio module changes are correct.  However, I think we can reduce ambiguity without compromising the meaning by changing "string of bytes" to "sequence of bytes" in both of those other cases.

Here is the updated patch.

In those contexts, bytes should not be linked to the bytes object.

Patch 4 looks fine to me.

Another uses of this idiom:

"string of 16 bytes" in uuid.rst and uuid.py
"string of *n* random bytes" in os.rst
"string of bytes" in bytesio.c

And a number of uses "bytestring", "byte string" and "bytes string". I think that either all occurrences should be changed consistently, or nothing. Personally I'm good with all these "string of bytes", "bytestring", etc, it is clear to me that that mean bytes or bytes-like objects. What is a bug, it's using the term "string" without "bytes" for bytes strings (there are a lot of such Python 2 remnants in the documentation and comments).

As I said, I think it's like fixing style issues in code: only do it if you touch the docs for some other reason.  But this patch is fine.  Although we should fix the 'string' when it is really 'bytes' issues, if we can find them.  Those are real bugs, not style issues.  But that should be a different issue or issues from this one.

New changeset 7ac8e8568da2 by R David Murray in branch '3.5':
#25916: fix a few 'string of bytes' references.
https://hg.python.org/cpython/rev/7ac8e8568da2

New changeset 089f7d761a9c by R David Murray in branch 'default':
Merge: #25916: fix a few 'string of bytes' references.
https://hg.python.org/cpython/rev/089f7d761a9c

Thanks SilentGhost.