As correctly pointed by drekbour :
JaCoCo performs analysis of bytecode - after compilation by Kotlin compiler
1.3.72 of
fun foo(inputStream: java.io.InputStream) {
val x = inputStream.bufferedReader()
}
output of javap
public static final void foo(java.io.InputStream);
descriptor: (Ljava/io/InputStream;)V
flags: (0x0019) ACC_PUBLIC, ACC_STATIC, ACC_FINAL
Code:
stack=4, locals=8, args_size=1
0: aload_0
1: ldc #9 // String inputStream
3: invokestatic #15 // Method
kotlin/jvm/internal/Intrinsics.checkParameterIsNotNull:(Ljava/lang/Object;Ljava/lang/String;)V
6: aload_0
7: astore_2
8: getstatic #21 // Field
kotlin/text/Charsets.UTF_8:Ljava/nio/charset/Charset;
11: astore_3
12: iconst_0
13: istore 4
15: aload_2
16: astore 5
18: iconst_0
19: istore 6
21: new #23 // class
java/io/InputStreamReader
24: dup
25: aload 5
27: aload_3
28: invokespecial #27 // Method
java/io/InputStreamReader."<init>":(Ljava/io/InputStream;Ljava/nio/charset/Charset;)V
31: checkcast #29 // class java/io/Reader
34: astore 5
36: sipush 8192
39: istore 6
41: iconst_0
42: istore 7
44: aload 5
46: instanceof #31 // class
java/io/BufferedReader
49: ifeq 60
52: aload 5
54: checkcast #31 // class
java/io/BufferedReader
57: goto 71
60: new #31 // class
java/io/BufferedReader
63: dup
64: aload 5
66: iload 6
68: invokespecial #34 // Method
java/io/BufferedReader."<init>":(Ljava/io/Reader;I)V
71: astore_1
72: return
clearly shows that there is a branch - see offsets 46 and 49 above.
Compilation of
fun foo(inputStream: java.io.InputStream) {
val temp = inputStream.reader(); val x = if (temp is
java.io.BufferedReader) temp else java.io.BufferedReader(temp,
DEFAULT_BUFFER_SIZE)
}
produces almost the same bytecode.
AFAIK there are no markers in bytecode to realize that
kotlin.internal.InlineOnly function was inlined.
So there is no way for JaCoCo to distinguish whether bytecode was produced
from source code as in the first example or as in second.
On Sunday, April 26, 2020 at 11:53:43 PM UTC+2, Giacomo Boccardo wrote:
> I saw that, but I think that, without modifying the function, I cannot
> cover the "(this is BufferedReader)" condition because neither I can force
> that branch from the argument "inputStream" nor I can mock inline functions.
Again as correctly pointed by
On Monday, April 27, 2020 at 12:39:50 PM UTC+2, [email protected] wrote:
> One of things I've found with higher-level language features in the JVM
> are instances of this impossible-to-cover scenario (I seem to remember Java
> try-with-resources is/was much the same). *It's really a "limitation" of
> the compiler* not inferring that condition has only one possible outcome,
> *not
> of Jacoco* for correctly interpreting the bytecode it saw.
Existence of such unreachable bytecode is bad not only for code coverage
measurement but also for the size of class files and performance. So please
report this to the developers of Kotlin. Either standard library functions
or compiler should be improved to not produce such unreachable bytecode. Or
for JaCoCo there should be a way to identify bytecode produced by
kotlin.internal.InlineOnly.
Regards,
Evgeny
--
You received this message because you are subscribed to the Google Groups
"JaCoCo and EclEmma Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To view this discussion on the web visit
https://groups.google.com/d/msgid/jacoco/3750b722-5bba-40f2-a388-6c6d72a72530%40googlegroups.com.
