adriaanjacobs

Thanks for the quick response. Changing the effect worked for me, thanks. AFAIK, the loading of dynamic libraries at run-time through mechanisms like `dlopen` does not involve any special system...

LLVM 15 still fully supports typed pointers (ref), and using the deprecated interface is only a warning, so adding ``` add_compile_options(-Wno-deprecated-declarations) ``` to the CMakeLists.txt worked perfectly fine for me...

Another quick question: it looks like all functions that are annotated with STATIC in extapi.c are also annotated with ALLOC_RET; is that intentional? The aliasing behaviors of those annotations seem...

> Though calling the same static function, it will return different new objects. You could have a try to validate it. > > If a function in extapi.c has one...

> For a function that is STATIC, both annotations should be removed and make the function body return a malloc object so that the application code will share the same...

I see. From the point of view of allocation wrappers, we frequently encounter not simply wrapper functions (such as xmalloc), but also initialization functions or functions that contain additional pointer...

> In your case, you can simply remove the ALLOC_RET annotation and use your self-defined function body in ExtAPI.c It should work. Okay, but would SVF not treat all calls...

Proposal: would it be possible to "inline" the PAG information obtained from the extapi entries in their callers (1-level context sensitive)? Most extapi functions could then be entirely defined by...

re 1: I'm not too familiar with how the SVFIR is constructed, but I believe calls to functions that have an extapi counterpart should be handled specially. When encountering a...

> I agree with your idea of inline the PAG information from the external API into the caller's PAG. However, constructing the PAG can involve various data structures, and if...