Thanks django at kerz.id.au.

Also alleviate issues with weird file names typically found in
MEDIA_ROOT directories (#23010).
Thanks Tim Graham for the review.

Thanks melinath for the report.

Thanks semenov for the report and Florian for investigation.

Mention of custom user models has been removed since UserCreationForm
didn't support custom user models anyway. Refs #19353.

This fixes a regression introduced in 6d302f63.
Thanks lorinkoz at gmail.com for the report, Collin Anderson
for the initial patch and Simon Charette for the review.

This fixes the Chinese language issues described in #23005 but
also provides for other fallback exceptions by updating the
LANG_INFO structure.
Thanks caxekis at gmail.com for the report and Tim Graham for the
review.

Thanks bendavis78 for the report and Tim Graham for the review.

Generic Foreign Keys can't be used as lhs in lookups for historical
reasons. Django 1.6 gave a FieldDoesNotExist exception when using GFKs
as lhs in lookups, but due to regression caused by lookup refactor patch
(20bab2cf) the exception type was
changed to AttributeError.

It might be a good idea to add support for gfk__exact and gfk__in lookups
later on.

Thanks to glicerinu@gmail.com for the report. The code in this commit
was written by Tim Graham.

The reason for the regression was that the GenericForeignKey field isn't
something meta.get_field_by_name() should return. The reason is that a
couple of places in Django expects get_field_by_name() to work this way.
It could make sense to return GFKs from get_field_by_name(), but that
should likely be done as part of meta refactoring or virtual fields
refactoring patches.

Thanks to glicerinu@gmail.com for the report and to Tim for working on
the issue.

Moved the {% lorem %} tag to built-in tags.

Thanks generalov for the report.

Accounted for the three stages of population: app configs, models,
ready() methods of app configs.

The rate at which we've increased this has not been keeping up with hardware (and software) improvements, and we're now considerably behind where we should be. The delta between our performance and an optimized implementation's performance prevents us from improving that further, but hopefully once Python 2.7.8 and 3.4+ get into more hands we can more aggressively increase this number.

It's been marked as "not recommended" since 1.6. Anyone still using it
may refer to older versions of the docs.

Thanks to @timgraham for the review.