Kotlin main class name is "mangled" by kotlinc

[mpsijm]

Follow-up of #124.

While preparing a recent contest, we found out that Kotlin does not only capitalize the first letter of the file, but also renames the main class to make it a valid Java class name. For example: test-test.kt is compiled to Test_test.class (note how the hyphen turns into an underscore). See the script below for a couple of weird filenames that I tested with.

Script used for testing names

for name in test Test test-1 test-test Têśt 'test+&~$test'
do
    echo
    echo $name
    echo 'val x = 42' > $name.kt
    kotlinc $name.kt
    ls *.class
    rm -f *.class *.kt
done

$ ./test.sh

test
TestKt.class

Test
TestKt.class

test-1
Test_1Kt.class

test-test
Test_testKt.class

Têśt
error: source file or directory not found: T????t.kt
ls: cannot access '*.class': No such file or directory

test+&~$test
Test____testKt.class

The problem with this "name mangling" is that Kotlin does document that this happens, just not how exactly. We could add in a string replace that turns every non-alphanumeric character into an underscore (_), but that only solves part of the problem: there might be characters that get a different treatment than just turning them into underscores, and files with non-ASCII characters apparently don't even compile (e.g. Têśt.kt).

One thing I did figure out: kotlinc dumps the name of the compiled class in META-INF/main.kotlin_module, so we could fetch it from there, but I'm not sure how that would fit in the problemtools pipeline.