Review com/reflect.d on what to do about only receiving one name

[JesseKPhillips]

The following link describes a problem and solution, it should be reviewed in the future.

http://www.dsource.org/forums/viewtopic.php?p=22965&sid=c93941193d22f5f7c9d5b2f11302a2d3#22965

[Numpsy]

Thinking about this now, the '0xbaadfood' thing that i mentioned in that post might just be because the memory allocated by CoTaskMemAlloc() on line 1388 of reflect.d is never initialised. That memory should probably be zeroed before use.

Changing line 1397 from bstrNames[0] to bstrNames[1] might also fix it (by always setting the parameter name to 'value'). Would need to test that though.

[Numpsy]

On a related note, there might be a memory leak in getParameters() -> line 1394 frees the first element of the name array, but i think it might need to free all the members that were set by GetNames()

[Numpsy]

I've changed the code in reflect.d a bit so that unnamed params to property functions should be left null, and will then be set to the user defined 'unnamed' value when tlbimpd writes the D source.

[Numpsy]

Seeing this reminds me of something i thought of ages ago and forgot about: Is there actually any need to allocate an array with CoTaskMemAlloc instead of just using a normal D array? Would avoid the need to tidy up manually, and possibly the memset as well?

I also thought that the call to freeBstr(bstrNames[0]) was needed because the later calls to fromBstr that free the contained BSTRs only free bstrNames[1]+, but i'm a bit hazy on the area.

[JesseKPhillips]

I was considering using the stack, but wanted to leave as much the same as I don't have good testing for these things yet.

The freeing call is CoTaskMemFree() which is a Windows function call and should be freeing all elements. It is after all meant to be used on memory allocated by the CoTaskMemAlloc() function.

[Numpsy]

I mean that the call to typeInfo_.GetNames() allocates some BSTRs that i think you have to manually free with SysFreeString. (the calls to fromBstr in the following loop call SysFreeString on their parameters, but that is never called for the first element in the array).

[JesseKPhillips]

You are correct, I'll have to put that back and consider a way to make those allocations and their release obvious.