Thanks to Collin Anderson, Tim Graham, Gabe Jackson, and Marc Tamlyn
for their input, ideas, and review.

Refactored bump_prefix() to avoid infinite loop and allow more than
than 5 subquires by extending the alphabet to use multi-letters.

refs #22340.

refs #22340.

The subclass check in ModelAdmin.get_formsets_with_inlines() wasn't
tested correctly because of the super() call in
EpisodeAdmin.get_formsets().

And removed Ignore*DeprecationWarningsMixin, now obsolete.
Thanks Berker Peksag and Tim Graham for the review.

This reverts commit a2e3c969.

The deprecation was moved back to 1.9 in
61da5f3f.

refs #22563

Thanks john_scott for the report and Markus Holtermann for review.

Refs #23271

Thanks to Michael D. Hoyle for the report and Tim Graham for the review.

Thanks Daniel Hahler for the report and Tim Graham for the review.

Thanks kabakov.as@gmail.com for the report, and Aymeric Augustin,
Simon Charette for the reviews.

Make sure PO catalog text fetched from gettext programs via standard
output isn't corrupted by mismatch between assumed (UTF-8) and real
(CP1252) encodings. This can cause mojibake to be written when creating
or updating PO files.

Also fixes #23311.

Thanks to contributor with Trac nick 'danielmenzel' for the report,
excellent research and fix.

All tests now rely on TEMPLATES.

This avoids leaving projects silently vulnerable when this option is set
to a string instead of a one-item tuple containing that string, a very
common misconfiguration.

Previously, when a template was rendered with RequestContext, inclusion
tags were rendered with a plain context, losing additional information
available in the RequestContext.

The (admittedly bizarre) implementation of RequestContext.new() has the
side-effect of not running template context processors, making this
change backwards-compatible.

dictionary and context_instance and superseded by context.

Refactored tests that relied context_instance with more modern idioms.