Open Toizi opened 6 years ago
can you share this binary? is the bug still there after so much time?
radare2 3.6.0-git 22086 @ linux-x86-64 git.3.5.1-263-g762b8e2d5
commit: 762b8e2d50ffff49354511dab0203cdf8187c84f build: 2019-06-17__09:16:05
This version (should be current master) still has this issue.
As I said in the bug report, the binary can be found in the r2 regression repo radare2-regressions/bins/jmptbl/ls_amd64
works fine for me
@radare Maybe my bug report was not worded clearly enough.
At address 0x004020a8
- which is part of the relocations, i.e. data - the bytes represent a call
when disassembled, which makes radare create this function from your screenshot. However the relocation data should not be disassembled because it's not actually code. In this case, it is not just some wrong xref but also the function is broken because of this. To look at the broken function, pdf @ 0x00407f80
and scroll down until you can see an invalid opcode. Right after the invalid, the function from your screenshot starts, which is not actually a valid function for a compiler generated binary.
Hope this clears things up
Work environment
Expected behavior
There are no call xrefs
Actual behavior
There is a call xref into the middle of a function, messing up disassembly
Steps to reproduce the behavior
The binary is in the r2 regression repo
radare2-regressions/bins/jmptbl/ls_amd64
I have looked into the code already and the issue is probably somewhere in
libr/core/canal.c r_core_anal_search_xrefs
, where it disassembles anything looking for jumps/calls@sivaramaaa