- Validate filename returned by FileField.upload_to() not a filename
passed to the FileField.generate_filename() (upload_to() may
completely ignored passed filename).
- Allow relative paths (without dot segments) in the generated filename.
Thanks to Jakub Kleň for the report and review.
Thanks to all folks for checking this patch on existing projects.
Thanks Florian Apolloner and Markus Holtermann for the discussion and
implementation idea.
Regression in 0b79eb36915d178aef5c6a7bbce71b1e76d376d3.
Address a long standing bug in a Where.add optimization to discard
equal nodes that was surfaced by implementing equality for Lookup
instances in bbf141bcdc31f1324048af9233583a523ac54c94.
Thanks Shaheed Haque for the report.
The 'db' and 'passwd' connection options have been deprecated, use
'database' and 'password' instead (available since mysqlclient >= 1.3.8).
This also allows the 'database' option in DATABASES['OPTIONS'] on MySQL.
- Replaced datetime.utcnow() with datetime.now().
- Replaced datetime.utcfromtimestamp() with datetime.fromtimestamp().
- Replaced datetime.utctimetuple() with datetime.timetuple().
- Replaced calendar.timegm() and datetime.utctimetuple() with datetime.timestamp().
In Python 3.9.5+ urllib.parse() automatically removes ASCII newlines
and tabs from URLs [1, 2]. Unfortunately it created an issue in
the URLValidator. URLValidator uses urllib.urlsplit() and
urllib.urlunsplit() for creating a URL variant with Punycode which no
longer contains newlines and tabs in Python 3.9.5+. As a consequence,
the regular expression matched the URL (without unsafe characters) and
the source value (with unsafe characters) was considered valid.
[1] https://bugs.python.org/issue43882 and
[2] 76cd81d603
The validate_file_name() sanitation introduced in
0b79eb36915d178aef5c6a7bbce71b1e76d376d3 correctly rejects the example
file name as containing path elements on Windows. This breaks the test
introduced in 914c72be2abb1c6dd860cb9279beaa66409ae1b2 to allow path
components for storages that may allow them.
Test is skipped pending a discussed storage refactoring to support this
use-case.
Having lookups group by subquery right-hand-sides is likely unnecessary
in the first place but relatively large amount of work would be needed
to achieve that such as making Lookup instances proper resolvable
expressions.
Regression in 35431298226165986ad07e91f9d3aca721ff38ec.
Thanks James A. Munsch for the report.
Subquery deconstruction support required implementing complex and
expensive equality rules for sql.Query objects for little benefit as
the latter cannot themselves be made deconstructible to their reference
to model classes.
Making Expression @deconstructible and not BaseExpression allows
interested parties to conform to the "expression" API even if they are
not deconstructible as it's only a requirement for expressions allowed
in Model fields and meta options (e.g. constraints, indexes).
Thanks Phillip Cutter for the report.
This also fixes a performance regression in bbf141bcdc31f1324048af9233583a523ac54c94.
Thanks Zain Patel for the report and Simon Charette for reviews.
The exception introduced in 6307c3f1a123f5975c73b231e8ac4f115fd72c0d
revealed a possible data loss issue in the admin.
Regression in 3a505c70e7b228bf1212c067a8f38271ca86ce09.
Nonlitteral right-hand-sides of lookups need to be wrapped in
parentheses to avoid operator precedence ambiguities.
Thanks Charles Lirsac for the detailed report.
This issue started manifesting itself when nesting a combined subquery
relying on exclude() since 8593e162c9cb63a6c0b06daf045bc1c21eb4d7c1 but
sql.Query.combine never properly handled subqueries outer refs in the
first place, see QuerySetBitwiseOperationTests.test_subquery_aliases()
(refs #27149).
Thanks Raffaele Salmaso for the report.
