IndirectAnalyzer: fix potential segfault

[ilil01]

When processing a call instruction, the IndirectControlFlowAnalyzer::FindAllThunks method assumes that both the instruction and its target are located in the same Region. This is not always the case — for example, in 32-bit libc there are instructions in __libc_freeres_fn section calling functions from .text. In such cases this method provokes a segmentation fault since the result of the getPtrToInstruction method is not tested against nullptr.

[bbiiggppiigg]

If the fix is just adding a if(target) check around the decoder and the for loop, it looks reasonable to me.
However, the indentation seems to be a bit off when viewed in the browser.

[ilil01]

If the fix is just adding a if(target) check around the decoder and the for loop, it looks reasonable to me. However, the indentation seems to be a bit off when viewed in the browser.

Yes, the fix is exactly that. What about the indentation, it's inconsistent in this whole function, especially due to mixed use of tabs and spaces. I decided to also make the indentation more consistent, at least in the changed scope. I can revert
indentation changes if that is more desirable.

[kupsch]

@ilil01, can you open a bug report issue for this problem with the information on how to reproduce? I am unable to reproduce this with a simple test. This will also allow us to make a more robust fix in the future.

[ilil01]

@ilil01, can you open a bug report issue for this problem with the information on how to reproduce? I am unable to reproduce this with a simple test. This will also allow us to make a more robust fix in the future.

I've opened issue #1680

[kupsch]

This prevents the crash. Fixing issue #1680, will be a more complete fix to find the correct region and instruction.