Fix mixed UTF-8 Java and Kotlin builds (#178)
When building .java files using the native Java rules, the environment
variable LC_LANG=en_US.UTF-8 is explicitely set for the actual build
action, see: https://git.io/fhhc7
When building a mixed Kotlin + Java module on Linux, the .java sources
are not handled by the native Java rules, and as a result are treated as
encoded using US-ASCII due to sandboxing which strips the environment
from all locale information, causing the compiler to default to the
'C' locale.
This patch adds this environment variable when building Kotlin + Java
modules and has no side effects on Kotlin-only modules as UTF-8 is the
only valid encoding for .kt source files.
diff --git a/kotlin/internal/jvm/compile.bzl b/kotlin/internal/jvm/compile.bzl
index dc08928..74e3ad6 100644
--- a/kotlin/internal/jvm/compile.bzl
+++ b/kotlin/internal/jvm/compile.bzl
@@ -231,6 +231,9 @@
arguments = [args],
progress_message = progress_message,
input_manifests = input_manifests,
+ env = {
+ "LC_CTYPE": "en_US.UTF-8" # For Java source files
+ },
)
return struct(
