FWD » Core Development » Base Language

This is a regression of #9448, where BinaryData.getMethodSetup now returns false if the size is 0. This is incorrect for memptr generated by native access, as FWD can't tell the buffer sizes of pointers generated by arbitrary libraries. Currently, the native invoker is completely broken for API that return pointers, as NativeBuffer.restore does something like:

memptr local = new memptr();
local.setPointerValue(addr);
val = local.getInt64(1);

The getInt checks for getMethodSetup which fails due to size being 0. So, the getter returns ?.

I have a fix I will post immediately.

I committed 11042a/rev. 16324.

• I forced the size of the memptr to be 4 or 8 to represent the pointer gathered through native interface.
• The goal of the code altered is to retrieve the data at the provided memptr, which is expected to be a pointer (as NativeBuffer is the governing class, not @NativePrimitive).
• That native pointer will be eventually assigned to the memptr. This is sent back to the server and the input-output parameter is assigned properly. At that point, we can't find out the length of the native pointer, as it will most probably refer a C struct that is arbitrarily large. That pointer is usually sent back to other native API.

This unblocks the testing of #10981 completely.

Code Review Task Branch 11042a Revision 16324

No objections. The change seems right.

Tested 11042a with two unit test suites (one from #9448 that caused the regression). There were no issues found. The POC that required this fix is properly working.
Preparing to merge 11042a to trunk.

Branch 11042a was merged into trunk as rev. 16340 and archived.