The code generation uses a mixture of literal `\n` characters and `appendln`. The latter insert `\r\n` on Windows by default, causing generated files to contain a mixture of line endings.
This commit sets the `line.separator` system property for the generator to `\n` so that `appendln` will never insert `\r` characters. As an additional measure, `.gitattributes` files were added to checkout generated stdlib files always with LF line endings.
- Allow to select target platform and target directory to generate source for
- Put copyright profile to resources to have it in the resulting jar
- Output error for missing bodies, but do not stop generation on the first error
- Add Kotlin/Native platform.
- Rename classes.
- Introduce TemplateGroup and TemplateGroupBase to group template values.
- MemberBuilder: add support for more properties.
Required as a workaround for Proguard bug
https://sourceforge.net/p/proguard/bugs/664/
When processing bytecode generated by Kotlin compiler for constructor
call with stack spilling during arguments evaluation, ProGuard performs
an equivalent transformation for the bytecode, but emits invalid stack
frame information.
In JVM 1.6, such invalid stack frames are ignored and re-evaluated by
JVM during bytecode verification.
In JVM 1.8, such invalid stack frames cause VerifyError.
Build parameters (with corresponding project properties):
- build.number (buildNumber) - build number from build server, goes into manifest, by default snapshot
- deployVersion (kotlinVersion, project.version) - version of artifacts, by default build.number
- bootstrap.kotlin.version (bootstrapKotlinVersion) - version of bootstrap compiler
Jerome Prinet