Recent commits to derivepassphrase.git (bdf95bab1e404e34f6a68c5b2bf72d510cd23556)

Fix typing of the dummy vault configuration settings in the constructor

2025-11-27T21:59:27+01:00

One of our tests, a "manual sanity-check" kind of test, tries to use the
`tests.data.DUMMY_CONFIG_SETTINGS` symbol as keyword arguments in the
`vault.Vault` constructor.  This appeared to work with older `mypy`
versions, but now raises an error about extra keys in the symbol that
the constructor does not accept as keyword arguments.  (Because the
dummy config settings are actually typed as vault config "services"
settings, which *do* allow more settings than the constructor uses as
arguments.)

So, we now arrange for a second symbol with effectively the same
contents as the dummy config settings, but explicitly typed as
constructor arguments, to be used in the "manual sanity-check" test
instead.

(Implementation-wise, we introduce a new base type
`_types.VaultConstructorArgs`, and let
`_types.VaultConfigServicesSettings` inherit from that.  This is easier
than attempting to constrain the latter type after the fact.  We
introduce the new base type directly in the `_types` module, even though
it is otherwise only used in the tests, because the derived type (from
the `_types` module) also needs access during construction.)
</pre>

Add CPython 3.14 to the list of test environments

2025-11-26T21:02:16+01:00

By now, CPython 3.14 has been released, as has a more recent version of
`hatch` that knows where to obtain CPython 3.14.  It is thus viable to
officially test with CPython 3.14 with only standard `hatch` installs.
</pre>

Fix more broken tests on The Annoying OS

2025-11-26T20:50:58+01:00

The broken tests relate to SSH agent handling.  We now have a fake SSH
agent, and I mistakenly believed that because this agent is always
available, tests no longer needed to ensure agent accessibility; the
fake agent would always be available.  But that's not how the SSH agent
test fixture works: you still need to explicitly *request* the agent as
part of your setup.  Otherwise you will run into the usual "no agent
could be located" or "no agent is supported on this system" error
situations (which is exactly what was happening on The Annoying OS).
The only thing that really changed is that we now know that *some* of
the test fixture instances (the ones with fake agents) will not
auto-skip due to lack of support.
</pre>

Fix broken links in the documentation caused by renaming or splitting modules

2025-11-26T20:49:34+01:00

</pre>

Split the basic command-line tests, again

2025-11-26T20:46:42+01:00

(This is part 11 of a series of refactorings for the test suite.)

Split the basic command-line tests into four thematic groups, in four
separate test files:

  * tests for common subsystems of the `derivepassphrase` CLI, plus
    miscellaneous basic tests

  * basic tests for the `derivepassphrase vault` CLI

  * tests for the configuration management options of the
    `derivepassphrase vault` CLI

  * tests for the note handling machinery of the `derivepassphrase
    vault` CLI

Also update the documentation to reflect the module split.
</pre>

Fix missing documentation in the test suite

2025-08-31T21:24:09+02:00

Update the documentation (files, navigation) to reflect the test files
split and the introduction of the `tests.data` and `tests.machinery`
packages.  Additionally, turn the package descriptions in the `tests`
module docstring into proper hyperlinks.
</pre>

Fix some documentation issues with the heavy-duty command-line inteface tests

2025-08-31T21:18:26+02:00

Fix both docstrings and inline commentary.
</pre>

Refactor the heavy-duty command-line interface tests

2025-08-31T21:09:13+02:00

(This is part 10 of a series of refactorings for the test suite.)

In the heavy-duty command-line tests, factor out common test setups and
common hypothesis strategies.  In particular, share the hypothesis
strategies and the skeletal procedure for both affected state machines
because they offer more or less the same transition rules.

For the hypothesis test machinery, fix some types and some import
aliases that affect the heavy-duty command-line tests.
</pre>

Refactor the "all CLIs" command-line interface tests

2025-08-29T20:00:38+02:00

(This is part 9 of a series of refactorings for the test suite.)

In the "all CLIs" tests for the command-line interface, factor out the
common test operation for the "help text" and "version output" tests.
Both sets of tests are identical for all command-line interface entry
points, save for the command-line and the expected lines/data in the
output.

Furthermore, use a helper function to make the `KnownLineType` enum
definition more pleasant to read.
</pre>

Refactor the basic command-line interface tests

2025-08-29T19:45:20+02:00

(This is part 8 of a series of refactorings for the test suite.)

In the basic tests, factor out the common test operation or the common
environment setup for each group of related tests, whichever is more
feasible.  For some groups of related tests, if there were similar tests
that differed only in details or in parametrization data, combine them
under a new parameter set, if necessary.

Both the test file and the commit should be further split up, but I am
not yet sure how.
</pre>