- update unescaping from \u{…} to \uXXXX in regex
- re-unescape in the string analyzr (empty-compiler.ts) so that we catch PUA etc.
- fix some XML files that had bad DTDs
- common; updates to string util api
#10319
- when a required section is missing, only throw an internal error IF we haven't already flagged a problem. for example, if 'vars' fails to compile because of a problem, and we already have an error on file, then don't throw an internal error that 'bksp' depends on uninitialized 'vars'.
- mimic this flow in the test helpers.
- add a test case for this in the test helpers, in test-tran
- add a strictError flag to say that no additional errs are allowed.
Fixes: #10291
Fixes#10150.
This feature adds a command line parameter to kmc to control the
severity of INFO, HINT and WARN level messages. The severity of any of
these messages can be reduced all the way to INFO, or even disabled, or
increased to ERROR. The compiler and the reporter will honour the new
code and fail builds when failure-level codes are encountered.
Inputs are validated against all the available compiler messages in all
modules. This also provides a good pathway for automated future
documentation generation of compiler messages.
In order to process the input parameters cleanly, have refactored some
of kmc's build command logic.
This commit increases code coverage for kmc to roughly 50%.
`outFile` is not a compiler option, and so it should not be a part of
the `CompilerBaseOptions` interface. This formally separates the
command-line options from the compiler options, which eliminates a
number of restrictive structures.
Part of #9473.
KmnCompiler now implements KeymanCompiler, including the returned
artifacts. Also establishes the common types for the compiler interfaces
and consolidates and renames various API surfaces for kmc-kmn.
- also found that kmc's validate() should have been async but wasn't, a little bit of churn because of this
- unassigned not implemented yet, but tests in place for it.
- change SectionCompiler.postValidate to only be called during the validate() run of compilation. This way messages will only show once.
For: #9446
Relates to #9948.
While the schema validation does check the version already, it gives a
fairly obtuse message, and we can do better.
Found a secondary issue with a iterator type guard needed in
kpj-file-reader when preparing unit test, which was probably masked in
other tests by not verifying all messages generated, so updated all
infrastructure message tests to verify the full set of messages
generated.
