Double slashes when generating the inventory file URL

[replaceafill]

Describe the bug

This might be considered an edge case but I've noticed that if the value tuple of a intersphinx_mapping contains a base URI with a trailing slash and None as the inventory file, Sphinx will try to load a URL with double slashes.

How to Reproduce

Using Sphinx 8.0.2 set the following in the conf.py file of a sphinx-quickstart project:

extensions = ["sphinx.ext.intersphinx"]
intersphinx_mapping = {'python': ('https://docs.python.org/3/', None)}

Run make html and notice the double slashes in the output message:

...
loading intersphinx inventory 'python' from https://docs.python.org/3//objects.inv ...
...

The double slashes URL could be problematic depending on the web server setup.

Environment Information

Platform:              linux; (Linux-6.5.0-45-generic-x86_64-with-glibc2.35)
Python version:        3.10.12 (main, Jul 29 2024, 16:56:48) [GCC 11.4.0])
Python implementation: CPython
Sphinx version:        8.0.2
Docutils version:      0.21.2
Jinja2 version:        3.1.4
Pygments version:      2.18.0

Sphinx extensions

["sphinx.ext.intersphinx"]

Additional context

I think the problem could have been introduced in #12087 where:

sphinx/sphinx/ext/intersphinx/_load.py

Lines 185 to 186 in f90b506

	if not inv:
	inv = posixpath.join(uri, INVENTORY_FILENAME)

was changed into:

sphinx/sphinx/ext/intersphinx/_load.py

Line 214 in fadb6b1

inv = f'{project.target_uri}/{INVENTORY_FILENAME}' if location is None else location

posixpath.join removes the trailing slash in the target URI, but the f-string doesn't.

[AA-Turner]

PR welcome!

A

[picnixz]

Ah yes, this one is on me. I don't remember why I did this. Maybe I incorrectly thought that the target_uri was stripped of its slash and wanted to reduce the imports. If this is the case, then I'm sorry. I'll do something for that this w-e or someone can fix it (it shouldn't be that hard).

[jayaddison]

Would removesuffix('/') be sufficient to fix this, or should we reinstate posixpath joining?

[picnixz]

Technically, since we are dealing with URLs, we should have used rstrip('/') because if you put 3 forward slashes, then

>>> posixpath.join('https://a.com//', 'b')
'https://a.com//b'

I'll prepare a PR.

[jayaddison]

Websites probably shouldn't accept significant-multiple-separators in their paths, but I think that technically it's valid for them to do that. (serving different content at foo.invalid/file.txt from foo.invalid/////file.txt, for example).

But foo.invalidobjects.env is never what we want, so I wonder whether the logic here should in fact be base + ('/' if not base.endswith('/') else '') + inv_filename.

[jayaddison]

(and then it's a question of whether we do that ourselves or via posixpath -- as presented -- or some other mechanism. it's not a file path, so posixpath seems a bit odd to me, but in this use case, the relevant logic does probably overlap fairly closely..)

[picnixz]

I revert it to the old behaviour in #12807. I don't want to mess with URLs and get complaints later :'). If we want to move from our current logic, more places need to be changed (there are some ifs where we test whether the URI is in some set of URIs where we explicitly add a forward slash, I don't remember where but I know it exists).