Some polymorphic calls are not recognizable about pharos HOT 5 CLOSED

Thanks for this bug report. It looks like the problem is related to the JSON exporter, not the actual Prolog results, which is good news, because detection problems are usually harder to fix than reporting problems. It will probably take us a while to fix this, but it's on the list now.

from pharos.

Thank you. This work is really meaningful.

from pharos.

Just adding additional notes about this bug to make it clearer what the problem is. The JSON results for ooex_vs2008/Debug/oo.exe report:

"Vcalls": [{ "call": "4116a2", "targets": [{ "ea": "411970" }]}]

But the Prolog results for the same file are:

finalResolvedVirtualCall(0x4116a2, 0x417824, 0x411970).
finalResolvedVirtualCall(0x4116bb, 0x417814, 0x4119f0).
finalResolvedVirtualCall(0x4116d5, 0x417814, 0x4118f0).
finalResolvedVirtualCall(0x4116ea, 0x417824, 0x411970).

So it appears that perhaps our JSON reporter only ever reports the first virtual call?

from pharos.

The code to pull out virtual function call targets examines call descriptors. which apparently do not account for multiple targets:

from pharos.

Indeed I got 52 finalResolvedVirtualCall's in another test (from different callsites) but only a single VCalls entry in the entire json file.

from pharos.