RFR: 8340205: Native linker allows MemoryLayout consisting of only PaddingLayout [v8]
Per Minborg
pminborg at openjdk.org
Tue Nov 5 14:23:30 UTC 2024
On Tue, 5 Nov 2024 14:12:42 GMT, Per Minborg <pminborg at openjdk.org> wrote:
>> Also, don't we already check the alignment before when calling `checkMemberOffset`? Why is it checked again here?
>
>> Also, don't we already check the alignment before when calling `checkMemberOffset`? Why is it checked again here?
>
> Good catch.
> What happens here if an element is already aligned, and `padding.byteSize() == element.byteAlignment()`? e.g. `MemoryLayout.structLayout(MemoryLayout.paddingLayout(4), ValueLayout.JAVA_INT)`?
This is an error that existed before this PR. On the main branch:
@Test
public void alignedByInitialPadding() {
Linker linker = Linker.nativeLinker();
var struct = MemoryLayout.structLayout(
MemoryLayout.paddingLayout(Integer.BYTES),
JAVA_INT);
var fd = FunctionDescriptor.of(struct, struct, struct);
linker.downcallHandle(fd);
}
would produce:
test TestLinker.alignedByInitialPadding(): failure
java.lang.IllegalArgumentException: Member layout 'i4', of '[x4i4]' found at unexpected offset: 4 != 0
at java.base/jdk.internal.foreign.abi.AbstractLinker.checkMemberOffset(AbstractLinker.java:235)
at java.base/jdk.internal.foreign.abi.AbstractLinker.checkLayoutRecursive(AbstractLinker.java:195)
at java.base/jdk.internal.foreign.abi.AbstractLinker.checkLayout(AbstractLinker.java:176)
at java.base/java.util.Optional.ifPresent(Optional.java:178)
at java.base/jdk.internal.foreign.abi.AbstractLinker.checkLayouts(AbstractLinker.java:167)
at java.base/jdk.internal.foreign.abi.AbstractLinker.downcallHandle0(AbstractLinker.java:98)
at java.base/jdk.internal.foreign.abi.AbstractLinker.downcallHandle(AbstractLinker.java:92)
at TestLinker.alignedByInitialPadding(TestLinker.java:160)
at java.base/jdk.internal.reflect.DirectMethodHandleAccessor.invoke(DirectMethodHandleAccessor.java:104)
-------------
PR Review Comment: https://git.openjdk.org/jdk/pull/21041#discussion_r1829438324
More information about the core-libs-dev
mailing list