Improves build script performance by:
* using built-ins wherever possible (e.g. string splitting)
* eliminating redundant code
* using absolute (to $KEYMAN_ROOT) rather than relative paths to avoid realpath
* removing unnecessary `npm run` calls
BEFORE | AFTER
-------------------------------|----------------------
time ./core/build.sh --help |
real 0m2.116s | real 0m0.874s
user 0m0.578s | user 0m0.198s
sys 0m0.984s | sys 0m0.289s
-------------------------------|----------------------
time ./web/build.sh --help |
real 0m3.523s | real 0m0.757s
user 0m1.166s | user 0m0.320s
sys 0m2.273s | sys 0m0.455s
-------------------------------|----------------------
time ./web/build.sh -d |
real 1m34.750s | real 0m59.974s
user 0m9.721s | user 0m6.284s
sys 0m19.652s | sys 0m13.202s
@keymanapp-test-bot skip
Relates to #2761.
kmc now supports passing in a folder and building all Keyman source
files found in the folder, even if a .kpj is not present in the folder.
If a .kpj file exists, `KeymanDeveloperProject/Options/Version` will be
checked to determine behaviour. If this field is missing or "1.0", then
the classical behaviour persists.
If the Version field is "2.0", the `Files` key is ignored, and .kmn,
.xml, .model.ts, and .kps files under `sourcePath` are enumerated and
added to the in-memory project for build.
If a .kpj is not present in the folder, a virtual project is constructed
in memory with the following defaults:
* `version`: `2.0`
* `buildPath`: `$PROJECTPATH/build`
* `sourcePath`: `$PROJECTPATH/source`
* `compilerWarningsAsErrors`: `false`
* `warnDeprecatedCode`: `true`
* `checkFilenameConventions`: `true`
It is still considered an error to include both a keyboard and a model
in the same file, but a version 2.0 project will not include the field
determining its type in `Options/ProjectType`; this will instead be
checked during build.
The call to `kmc` can reference a folder or a .kpj file. The .kpj file
must have the same basename as the folder name (e.g.
`khmer_angkor/khmer_angkor.kpj`).
WARNING: this does not add support for 2.0 projects to Keyman Developer;
it is likely that opening a version 2.0 project in Developer will
currently corrupt it.
WARNING: it is not anticipated that kmcomp.exe will be updated to
support version 2.0 projects -- rather kmcomp.exe will be removed before
release of v17, and replaced with kmc.
Adds support for building .kpj projects. Required minor tweak to kmcomp
to allow it to build both .kmx and .js from .kmn command line.
Currently, calls kmcomp.exe to build .kmn files, until we support a
cross-platform kmc-kmn C++ project.
- spec: change caps value to 0x100 #7533
- add a 'mod_all' for validating valid modifier keys, plus tests
- add asserts on C++ side mapping modifiers to keyman constants
- use 'mod_all' to validate modifiers
- further fix modifier -> mod in binary
For: #7985 and related to #7533
- make ldml.cpp print FAILED on failure especially because of https://github.com/microsoft/vscode-cpptools/issues/487
- add vkey validation (#7135)
- add disp validation
- fix: a latent bug in disp! disp.count was === vkey.count, oops. Fix and test cases.
For: #7985
- add HardwareToKeymap in common/web/types
- add a mt-iso keyboard, very close to stock mt.xml
- add machinery for parsing modifier key sequences
- more test machinery
- add assertions against invalid modifier sequences
Fixes: #7965
- spec: update spec for layr.mod field
- core/include/ldm: update constants
- bitfield modelled after s_modifier_names[] / LCTRLFLAG etc.
- new test case k_004_tinyshift, also update k_010_mt
- update const builder to ignore all *_map properties
- split utility out to utils, add test
For: #7533
(cherry picked from commit 0346bea488)
- Apply suggestions from code review
- move m() into common code
- rename several error-related fields and types
- drop the 'unknown AJV error' - we expect errors to be populated
- halt compilation if validation fails, don't press on
For: #7534
Co-authored-by: Marc Durdin <marc@durdin.net>
(cherry picked from commit 05cc200f2027ee3c61280812a2de85ef20ea8b51)
- more refactor into common/web/types
- refactor test harness
- use reportMessage instead of throwing
For: #7964
(cherry picked from commit a112b847c4)
- use hexy and git-diff to show a rich diff when files are changed
- add additional round trip tests
- initialize the underlyingLayout to undefined to promote round tripping
- fix an issue with the CompilerCallbacks in the xml tests (broken in #8107)
- fix an issue with the kvk file writer where it should set strings irregardless (#7955)
Fixes#7967, again