Ensures we get a canonical tag per langtags.json as far as we possibly
can. This is a breaking change for the compiler as tags which were
formerly regarded as canonical are no longer regarded that way. This
mostly relates to script subtag but a secondary bug meant that some
other tags would have lost data in the canonicalization process
(because we did a lookup based only on the language subtag previously,
which is a no-no).
See keymanapp/keyboards#1452 for related work.
This improves the implementation for normalizing language tags
when installing a keyboard. Based on the Windows implementation
in `Keyman.System.CanonicalLanguageCodeUtils.pas`.
One difference to the Windows implementation is that the mapping of
3- to 2-letter tags is based on `langtags.json` instead of
`iso639-3.tab`.
Closes#3399.
Fixes#4280.
Adds checks to verify that `if()`, `platform()`, `baselayout()` and
`nul` are at the start of the context in the appropriate order.
Adds unit tests to validate these checks.
These are conditions that have been present in earlier versions of
Keyman but not enforced until now. Keyboards that do not meet these
conditions would not work correctly in all circumstances and should be
updated to meet the requirements.
An alternative would have been to reorder the context string but that is
much more complex as manipulation would also have been required for the
output string. The enforced order is logical and reduces confusion in
any case.
Updated documentation coming along shortly.
Fixes#4250.
This implements @jahorton's suggestion of including the original name
in a comment alongside the munged name, as well as appending the unique
integer identifier of each symbol.
Note: the error reporting code needed a massage to support Unicode error
strings; in the future we should update the C++ compiler to emit
Unicode error strings also.
Fixes#4393.
Fixes KEYMAN-DEVELOPER-46.
Fixes KEYMAN-DEVELOPER-45.
This arose because deleting a tab caused Delphi VCL to switch to another
tab in the page control, which is IMHO a bug in the TPageControl VCL
component. Furthermore, when the active page is changed under these
circumstances, the normal OnChange and OnChanging events are not fired,
which means that the Source tab (which is the lucky tab to be activated)
never gets initialised. This cascades eventually to a crash due to the
Source tab's text editor not being properly loaded.
Two parts to this fix:
1. Restore the active page when we delete a wordlist tab (RSP-32327);
this does not resolve the crash but does avoid weird side-effect of
the active tab changing.
2. Initialise the source tab when we load the form so that the
initialisation issue described above is avoided.
A separate fix in modelTsProjectFile avoids a (handled) null variant
conversion exception.
