The test .kps for binary_kvk_file tests was not very coherent, which
meant that it failed with the changes and additional consistency
checking included in this PR.
When compiling a package, rewrite the font data in embedded .kvk files
to use the facename from the font referenced in OSKFont or DisplayFont.
This also does some safety checking for cross-references of the .kvk,
and the font data.
Fixes: #12912
Adds a new design-time font property for when `&displayMap` is in use in
a keyboard, so that the web debugger can use that font as appropriate.
If the keyboard does not have a `&displayMap` store, then the DisplayMap
font property is disabled, and the web debugger will use the OSK Font
instead, as previously was the case.
Note that the font for `&displayMap` has to be installed on the keyboard
author's system in order for it to be served up.
Fixes: #12123
The font collection code was somewhat wrong in kmc-keyboard-info. It
collected all fonts referenced in the package, even for multi-keyboard
packages, which meant that the .keyboard_info file listed all fonts for
all languages. Furthermore, if a font was referenced in multiple
language entries in the .kps, then it would be repeated for each
language in the .keyboard_info. This patch addresses both of these bugs.
The good news is that this makes some of the .keyboard_info files
smaller. In particular, fv_all.keyboard_info goes from 4675 lines down
to 695 lines!
Fixes: #12852
Some additional items arose from this:
* Adds a command-line-tests unit test which does some basic sanity tests
on `kmc`.
* Cleans up some of the error reporting infrastructure in `kmc` to
ensure that thrown errors are handled better.
* Removes redundant subshell from `builder_run_action`.
Fixes: #12846
Originally reported in #12202. Check was added in #12204, backported to
17.0.329 in #12207, but #12597 reports this for 17.0.331. So added test
to verify that it is working as expected (it is).
Note that this only applies when using the `--for-publishing` build flag
in kmc, because it is checked in the kmc-keyboard-info stage. Guessing
this is where the confusion arose.
Consolidated tests which used a common pattern for kmc-keyboard-info
calls. The remaining tests do stubs or calls into internal functions, so
they cannot be easily deduped.
Fixes: #12597
- use path.join instead of path.resolve
- rename importsPath to cldrImportsPath
- remove k_015_importlocal from core, as it duplicated tests in common and developer
Fixes: #10649
This went into kmc-keyboard-info because we already do a lot of
processing there against langtags.json, whereas we do not currently have
langtags.json in kmc-package.
Adds an associated unit test.
Fixes: #12752
