Re: [Qemu-devel] [PATCH] libvixl: Add gcc format attribute

[Stefan Weil <sw@weilnetz.de>]

Am 18.06.2014 17:19, schrieb Richard Henderson:
>
> On 06/17/2014 09:45 PM, Stefan Weil wrote:
>>
>> Variable arguments usually are not converted to 64 bit values: if
>> they are smaller than int, they are expanded to int, and larger
>> values are passed as they are. But here obviously the compiler
>> expands uint32_t to int64_t. Why?
>
> They really really shouldn't be. It might be worth trying something
> more recent than 4.6.3, and if it persists file a bug. r~

I found the source of the problem.

The compiler is correct. Eight format strings in
disas/libvixl/a64/disasm-a64.cc are wrong.

The functions which are called don't come from
disas/libvixl/a64/assembler-a64.h as I expected.
They are generated by code in disas/libvixl/a64/instructions-a64.h:

  #define DEFINE_GETTER(Name, HighBit, LowBit, Func)             \
  inline int64_t Name() const { return Func(HighBit, LowBit); }
  INSTRUCTION_FIELDS_LIST(DEFINE_GETTER)
  #undef DEFINE_GETTER

So each of those functions really returns an int64_t which of course
should not use a "%d" format string.

Stefan