Fix the displacement offset for moffset-encoded operands

[jranieri-grammatech]

This was initially introduced in dce7da9 but lost in the LLVM 7 sync
in 5a99624.

[aquynh]

can you provide a testcase for this bug?

[jranieri-grammatech]

Sure!

Where should I add the test to? I've looked in tests/ and suite/ and nothing really stands out as a place to add a test that disassembles a few instructions and verifies their displacement offsets.

[aquynh]

can you put the code exposing this bug here?

[jranieri-grammatech]

#include <capstone/capstone.h>
#include <assert.h>
#include <string.h>

int main() {
  csh cs_handle;
  cs_err err = cs_open(CS_ARCH_X86, CS_MODE_32, &cs_handle);
  assert(err == 0);

  const char asm_bytes[] = { 0xA3, 0x00, 0x00, 0x00, 0x00 };

  cs_insn* insn;
  cs_option(cs_handle, CS_OPT_DETAIL, CS_OPT_ON);

  size_t count = cs_disasm(cs_handle, asm_bytes, sizeof(asm_bytes), 0, 0, &insn);

  assert(count == 1);
  assert(strcmp(insn[0].mnemonic, "mov") == 0);
  assert(insn[0].detail->x86.op_count == 2);
  assert(insn[0].detail->x86.operands[0].type == X86_OP_MEM);
  assert(insn[0].detail->x86.operands[1].type == X86_OP_REG);
  // The following assertions will fail without the bug fix.
  assert(insn[0].detail->x86.encoding.imm_offset == 0);
  assert(insn[0].detail->x86.encoding.imm_size == 0);
  assert(insn[0].detail->x86.encoding.disp_offset == 1);
  assert(insn[0].detail->x86.encoding.disp_size == 4);

  cs_free(insn, count);
  cs_close(&cs_handle);
}

[eschulte]

Is Capstone actively maintained? I've noticed numerous forks which seem to primarily exist to hold un-merged bug fixes.

For example at my company, we maintain both a persistent fork at https://github.com/GrammaTech/capstone/tree/next and a PyPi package at https://pypi.org/project/capstone-gt/.

I appreciate that Capstone is open-source software and this sort of maintenance can require significant uncompensated effort, so please don't take this as a complaint, but rather a request for information on Capstone maintenance plans so that we can plan our development accordingly. In particular, should we continue to bother opening issues and creating merge requests?

Thanks!

[kabeor]

Bug fix confirmed. Thanks for your contribution.
Merge now.