[next] SH: Use bitwise OR with mask for sign extension

[lhsazevedo]

Cherry pick of #2371 to next branch

[XVilka]

Though SH disassembler generates clang-tidy warnings:

/home/runner/work/capstone/capstone/arch/SH/SHDisassembler.c:102:9: warning: va_arg() is called on an uninitialized va_list [clang-analyzer-valist.Uninitialized]
                grp = va_arg(g, sh_insn_group);
                      ^
/usr/lib/llvm-14/lib/clang/14.0.0/include/stdarg.h:19:29: note: expanded from macro 'va_arg'
#define va_arg(ap, type)    __builtin_va_arg(ap, type)
                            ^~~~~~~~~~~~~~~~~~~~~~~~~~
/home/runner/work/capstone/capstone/arch/SH/SHDisassembler.c:100:9: note: Assuming 'n' is > 0
        while (n > 0) {
               ^~~~~
/home/runner/work/capstone/capstone/arch/SH/SHDisassembler.c:100:2: note: Loop condition is true.  Entering loop body
        while (n > 0) {
        ^
/home/runner/work/capstone/capstone/arch/SH/SHDisassembler.c:102:9: note: va_arg() is called on an uninitialized va_list
                grp = va_arg(g, sh_insn_group);
                      ^
/usr/lib/llvm-14/lib/clang/14.0.0/include/stdarg.h:19:29: note: expanded from macro 'va_arg'
#define va_arg(ap, type)    __builtin_va_arg(ap, type)
                            ^~~~~~~~~~~~~~~~~~~~~~~~~~
/home/runner/work/capstone/capstone/arch/SH/SHDisassembler.c:937:4: warning: Value stored to 'insn' is never read [clang-analyzer-deadcode.DeadStores]
                        insn = SH_INS_JMP;
                        ^      ~~~~~~~~~~
/home/runner/work/capstone/capstone/arch/SH/SHDisassembler.c:937:4: note: Value stored to 'insn' is never read
                        insn = SH_INS_JMP;
                        ^      ~~~~~~~~~~

cc @Rot127

[Rot127]

@lhsazevedo Please test it with clang-tidy-18. The CI Ubuntu only has version 14 or something. Which reports false positives for the va_args.
If everything is fine with clang-tidy-18 it can be ignored.

[lhsazevedo]

@Rot127 clang-tidy 18 also warns about uninitialized va_args. Perhaps it is a bug in clang-tidy?

/home/lhsazevedo/code/tweaking/capstone/arch/AArch64/AArch64Mapping.c:2162:14: warning: va_arg() is called on an uninitialized va_list [clang-analyzer-valist.Uninitialized]
 2162 |                 int Tile = va_arg(args, int);
/home/lhsazevedo/code/tweaking/capstone/arch/AArch64/AArch64Mapping.c:2211:18: warning: va_arg() is called on an uninitialized va_list [clang-analyzer-valist.Uninitialized]
 2211 |                 int8_t First = va_arg(args, int);
/home/lhsazevedo/code/tweaking/capstone/arch/SH/SHDisassembler.c:102:9: warning: va_arg() is called on an uninitialized va_list [clang-analyzer-valist.Uninitialized]
  102 |                 grp = va_arg(g, sh_insn_group);

There are more warnings for SHDisassembler, but they all seem to be unrelated to this change

[Rot127]

Sorry, those AArch64 ones can be ignored. I fixed them in the AArch64 update. Just check the SH one please.

[lhsazevedo]

@Rot127, I haven't had time to check the SH disassembler yet as the warnings are from a different part of the code. I'll get on it as soon as I can. Thanks for your patience!

[Rot127]

Little reminder :)

[lhsazevedo]

@Rot127 Done! Sorry for the delay. There were just two warnings, and one turned out to be a false positive.

---

P.S.: I noticed SH isn't listed in the auto-sync progress tracker. Any thoughts on how challenging its refactor might be?

[Rot127]

P.S.: I noticed SH isn't listed in the auto-sync progress tracker. Any thoughts on how challenging its refactor might be?

SH is not supported by LLVM. And Auto-Sync works exclusively on LLVM.
I could not find a fork which implements it. If you aware of one, please let me know and I take a look.

[Rot127]

@kabeor

[Rot127]

@lhsazevedo Ok, so @kabeor and me decided to suppress the warnings. Sorry for the unclear requests. Could you please reset to the previous commit and force push.

[lhsazevedo]

Of course, sure

[Rot127]

Mh, ok. So I think with the update clang-format-18 we get now some new stuff:

 /home/runner/work/capstone/capstone/arch/SH/SHDisassembler.c:1960:9: note: The value '29' provided to the cast expression is not in the valid range of values for 'sh_dsp_insn_type'
 1960 |                                                                 (sh_dsp_insn_type) SH_INS_DSP_PWAD,

I am sorry this PR takes so much longer although it is so simple. We try to improve code quality overall. If you could fix this issues there (multiple occurrences of casting an instr. enum to a type enum? No idea why, probably typos no one saw) I would really appreciate it.

[Rot127]

@lhsazevedo Ignore my comment above. I'll fix the issues in #2469
@kabeor This is ready to be merged.

[lhsazevedo]

Ok!

[kabeor]

Thanks for your contributions!