Instruction tracing can use wrong encoding mode #255
Assignees
Comments
arichardson
added a commit
that referenced
this issue
Jun 18, 2024
While this means we have to allocate a temporary buffer for the textual trace format, I don't see another workaround for the fact that we print the instruction disassembly after the instruction has executed and therefore has modified state (such as env->thumb) that affects the disassembly of instructions. Fixes: #255
arichardson
added a commit
that referenced
this issue
Jun 18, 2024
While this means we have to allocate a temporary buffer for the textual trace format, I don't see another workaround for the fact that we print the instruction disassembly after the instruction has executed and therefore has modified state (such as env->thumb) that affects the disassembly of instructions. Fixes: #255
arichardson
added a commit
that referenced
this issue
Jul 24, 2024
While this means we have to allocate a temporary buffer for the textual trace format, I don't see another workaround for the fact that we print the instruction disassembly after the instruction has executed and therefore has modified state (such as env->thumb) that affects the disassembly of instructions. Fixes: #255
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The problem we see here is that
arm_disas_set_infois called after all of the instruction side-effects have taken place, so we disassemble using the mode of the subsequent instruction rather than the current state.The only real workaround I can think of is eagerly expanding the disassembly and then storing that string in the trace buffer rather than storing the instruction opcode+size.
@qwattash what do you think about this approach?
The text was updated successfully, but these errors were encountered: