In the summary table at the top of library/functions, the links for frozenset and memoryview do not jump to the definition down in the same file, but link to stdtypes.

This is not a Sphinx glitch: that's where these classes are defined.

(Plus, dependencies work the other way round.)

memoryview and frozenset do have a small entry in library/functions, and the links at the top of the file are meant to jump to entries in the same file.

Well, then a different way of linking to them must be used.

Certainly.  I tried using `.memoryview` without success.  How would you do it?

You can put an anchor and use :ref:.

Sounds like the easiest solution that could work, thanks.  I’m on it.

I agree that all table entries should link to the entry in the same file.

dict() and set() also jump to CH. 4, in addition to memoryview and frozenset. Others all work properly.

The entries for classes should generally have a link to their section in '4. Built-in Types'. All do other than these:

range() definitely needs a forward reference to 4.6 sequence types 

bool() lacks forward references; it could have one to 4.2 Bool ops and 4.11.9 Bool values

object() -- no where to go.

slice() only has a reference to the glossary; it has no entry in Ch.4. "Slice objects have read-only data attributes start, stop and step which merely return the argument values (or their default). They have no other explicit functionality" is slightly wrong since there is the .indices method.

type(n,b,d) has no forward references as there is no where to go exactly

New changeset 7a05cb3beddf by Éric Araujo in branch '3.2':
Fix a few links in the table of built-in functions (#12298)
http://hg.python.org/cpython/rev/7a05cb3beddf

New changeset 58dd7addef3a by Éric Araujo in branch '3.2':
Add links from library/functions to other docs.
http://hg.python.org/cpython/rev/58dd7addef3a

> I agree that all table entries should link to the entry in the same file.
This is fixed in 3.2 and 3.3, with a bit of reST kludgery.  I’ll backport to 2.7 if there is no negative feedback.

> range() definitely needs a forward reference to 4.6 sequence types 
> bool() lacks forward references; it could have one to 4.2 Bool ops and 4.11.9 Bool values
I have added links in a subsequent commit.

> object() -- no where to go.
> type(n,b,d) has no forward references as there is no where to go exactly
I think this deserves another report.

> slice() only has a reference to the glossary; it has no entry in Ch.4. "Slice objects have
> read-only data attributes start, stop and step which merely return the argument values (or
> their default). They have no other explicit functionality" is slightly wrong since there
> is the .indices method.
Ditto.

Sandro ported to 2.7 in 07b3fc67bf45.  Thanks!

If I may make two remarks:

- Please duplicate commit messages from 3.2 and 2.7: it makes it easier to understand what a changeset is about without having to hunt.  (This does not apply to 3.3 as the changeset with the full commit message is a direct parent and thus easily found.)

- I’m not retired yet :)

Oh sorry Éric, I completely oversaw there was an issue associated with the cset - i'll pay more attention next time!

No problem, what counts is that our documentation and code get better for our users, not my ego :)