Thread safety problem in contrib.sitemaps

[Gabriel Hurley]

Since sitemaps are global objects, setting self.request on the object is thread-unsafe. ​Reported here by Justin Bronn.

[Gabriel Hurley]

Restores thread-safety to contrib.sitemaps

[Gabriel Hurley]

I believe this patch should restore thread safety by passing site as an argument rather than storing request on the Sitemap object. It is as backwards-compatible as possible, since prior to [13980] the get_urls method simply called Site.objects.get_current() directly, so this patch uses that behavior as a fallback for anyone who may have written custom view code which does not pass in a site as an argument.

[Luke Plant]

Thanks Gabriel. I will commit with changes:

• added tests for raising ImproperlyConfigured (and then added an import and some other logic that made the tests pass - hurray for TDD :-)
• used keyword arguments in to Sitemap.get_urls() instead of positional, in django/contrib/sitemaps/views.py and django/contrib/gis/sitemaps/views.py

I am unable to run the GIS testsuite (see #14439), so I have not checked that the GIS sections are correct, but they look correct are (hopefully) trivial.

[Luke Plant]

(In [14141]) Fixed #14433 - replaced a thread-unsafe solution to #10235 introduced in [13980]

This patch also addresses sitemap code found in contrib/gis, which [13980]
did not.

Thanks to gabrielhurley for the initial patch.

Refs #10235, #14386

[Luke Plant]

(In [14142]) [1.2.X] Fixed #14433 - replaced a thread-unsafe solution to #10235 introduced in [13980]

This patch also addresses sitemap code found in contrib/gis, which [13980]
did not.

Thanks to gabrielhurley for the initial patch.

Refs #10235, #14386

Backport of [14141] from trunk.

[Gabriel Hurley]

I knew the patch wasn't finished, but I had to run out the door. I wanted to give Justin or someone else who could run the GIS suite a chance to look at it. All the same, thanks for wrapping it up Luke.

[Jacob]

Milestone 1.3 deleted