User application request has an Authorization header so an Authorization
instance is returned but it can not be used for login.
Closes#12312
(grafted from 0a4e13728c0a9bb24536cb1033d95ca6749a2b9d)
--HG--
branch : 6.0
Since Werkzeug >= 2.3, it always tries to parse the Authorization header as a
digest. If the session value which is encoded in base64 needs padding, the
trailing `=` are interpreted as header values and the token attribute is not
filled.
In this case we must parse ourself the Authorization header to decode the
base64 value as session.
Closes#12259
(grafted from 0bd0b8a0e20ad8d2f3d13cc8d879c78717817b7a)
--HG--
branch : 6.0
As the cache is rollback to avoid dead lock, the transaction commit could
restore cache version that is no more valid.
For example the language cache of `ir.translation` which is filled initially
with only English and then reset when other languages are created by the `ir`
module. So at the end of `ir` module load, the cache is reset to contain only
English as it was the content before the first cache reset. This prevents to
load any other language from other modules.
So to ensure the next module is loaded using a fresh cache, all the caches are
cleared.
Closes#12246
(grafted from 5f14ca66744e53cd1d6200dfd569c2dca19f99c6)
--HG--
branch : 6.0
The werkzeug refactoring always return an Authorization instance.
If the type is our custom 'session', we must parse the token attribute.
Closes#12234
(grafted from 7fd0bb7e6864bdf3fa83fafb17262316fded5b0c)
--HG--
branch : 6.0
The sequence functions are using string parameters so for case sensitive
sequence name we must pass a string containing double-quoting if necessary.
Closes#12194
(grafted from ceb3118cd35511412d7cf8f2eae6c1891cf5f47a)
--HG--
branch : 6.0
When the reference field is validated (without check access), it should not
fail for SelectionValidationError if the user has no access to the resource.
Closes#12178
(grafted from 1bb024e29dfae1cbb0487b1f66d325f3bb0059fa)
--HG--
branch : 6.0
The tasks records are removed on rollback so there is no need to try to run
them.
(grafted from d2e0256f3478213d5c1ab375bc842c1222157692)
--HG--
branch : 6.0
For xxx2Many field, the contain test is not working correctly because the
field value is a tuple of records which is never in the list of invalid
relations.
Closes#12099
(grafted from 9dabdd3422a69e132269c4e9cc02076286aba3ed)
--HG--
branch : 6.0
The tasks list is global per thread. As the tests are run in the same thread,
we must clear remaining tasks that are not run by the test.
Closes#11978
(grafted from 2876d8571c510e7a1afd2c19da2dcebf4ebd0824)
--HG--
branch : 6.0
The left part must be returned because when the value is unique we also use the
name of field for further checks
Closes#11952
(grafted from 5ae26db33565f1d67e5eb2ae975f5c16a38cd4d9)
--HG--
branch : 6.0
If there was a next task planned very soon, this should not become the default
timeout for future loop.
issue11913
review415971003
(grafted from 10426538d63f736318c9eb5028cf34d423701422)
--HG--
branch : 6.0
Cédric Krier