Recent commits to derivepassphrase.git (bc5c69ba7b7181bbe656c4bb874f5bfa84233a2b)

Merge topic branch '2025-and-zlib-license' into master

2025-01-07T15:36:12+01:00

Note that this includes a change in the license for `derivepassphrase`.

* t/2025-and-zlib-license:
  Add changelog entry, and FAQ for "altered versions" marking
  Update copyright notices to 2025, and license to zlib/libpng
</pre>

Add changelog entry, and FAQ for "altered versions" marking

2025-01-07T15:35:45+01:00

</pre>

Update copyright notices to 2025, and license to zlib/libpng

2025-01-07T15:01:51+01:00

The MIT license was originally suggested by the package manager, for
compatibility with the broader ecosystem.  This surely makes sense for
software libraries, and for projects in general if you don't have
a strong stance on free/libre open source software licensing (beyond
copyleft vs. non-copyleft).  For `derivepassphrase`, I respectfully
disagree; my priorities in development and maintenance lie primarily
with ensuring quality software, and not having bugs attributed to me
that I did not cause.  The zlib/libpng license fulfills this need much
better than the MIT license, and has similar license compatibility with
other FLOSS licenses.  Thus the license change.
</pre>

Merge topic branch 'shell-completion' into master

2025-01-07T12:53:27+01:00

* t/shell-completion:
  Consolidate shell completion, add tests and fix Zsh output
</pre>

Consolidate shell completion, add tests and fix Zsh output

2025-01-07T12:10:27+01:00

We move the shell completion code in `derivepassphrase/cli.py` to a new
section and retire the specific handling of `VAULT_PATH` for the two
path completion functions.  Having actually tried this out interactively
in Bash, it does not work that well if we complete filenames, partial
filenames, and fixed strings in the same completion function: you would
have to set conflicting completion options for this function.  It *also*
does not work at all with `click`'s stock Bash completion script, which
would then discard either the fixed strings and partial filenames or the
complete filenames, depending on the order they are emitted in by our
shell completion code.  So abandon the completion handling of
`VAULT_PATH` in favor of standard filename completion, for paths.

Every supported shell has further limitations on which inputs it can
properly deserialize: Bash strips NUL characters from command
substitutions, and Fish (v3) breaks in most situations involving
newlines.  Furthermore, the stock shell completion scripts add their own
additional limitations: all shells use newline-terminated messages, so
embedded newlines in the completion item (type, value, help text) cause
(generally silent) parsing failures; and the Zsh completion functions in
particular parse completions as "name:description" pairs, and thus need
colons in the name to be escaped.

We fix the colon handling of the Zsh completion script by providing
a fixed serialization handler for Zsh.  We avoid all of the other
aforementioned issues by not returning any service names containing
ASCII control characters as completion items.  We also warn the user
upon importing or configuring such a service that its service name will
not be available for shell completion.

Finally, we document this new warning in the manpage and the completion
behavior in the changelog, and add tests for (the Python side of) the
completion machinery, dependent on the current serialization format.

References:
[fish-shell#10874](https://github.com/fish-shell/fish-shell/issues/10874),
[fish-shell#9693](https://github.com/fish-shell/fish-shell/issues/9693),
[fish-shell#751](https://github.com/fish-shell/fish-shell/issues/751),
[fish-shell#10651](https://github.com/fish-shell/fish-shell/issues/10651),
[fish-shell#9847](https://github.com/fish-shell/fish-shell/issues/9847),
[click#2703](https://github.com/pallets/click/issues/2703).
</pre>

Merge topic branch 'usage-fixes' into master

2025-01-07T11:13:37+01:00

* t/usage-fixes:
  Consolidate and slightly simplify some warning message emittances
  Add tests for help and version options, and color forcing
  Fix check of empty service name for `derivepassphrase vault`
</pre>

Consolidate and slightly simplify some warning message emittances

2025-01-07T11:09:38+01:00

</pre>

Add tests for help and version options, and color forcing

2025-01-01T14:08:48+01:00

These sections no longer need be excluded from coverage.
</pre>

Fix check of empty service name for `derivepassphrase vault`

2025-01-01T02:02:43+01:00

We asserted in c1bf00eadd1bf733ac25a25eafbe110d61936c54 that an empty
service name is both diagnosed, then treated as if it were missing later
on.  This wasn't actually strictly true: the checks for a missing or
superfluous service name tested for truthy service names instead of
non-missing service names, and the key selection and passphrase entry
dialogs happen before the check for a non-empty service name.  Fix these
two issues by checking for non-missing service names in the former case,
and moving the non-empty service name check further to the front in the
latter case.
</pre>

Merge topic branch 'no-color-force-color' into master

2025-01-01T00:57:42+01:00

* t/no-color-force-color:
  Add a changelog entry for the color suppressing/forcing support
  Support suppressing or forcing color output
</pre>