[kvm-unit-tests PATCH 1/6] libcflat: add PRI(dux)32 format types

[Alex Bennée]

Andrew Jones <drjones at redhat.com> writes:

> On Thu, Jan 12, 2017 at 01:29:24PM +0100, Paolo Bonzini wrote:
>>
>>
>> On 11/01/2017 17:28, Alex Bennée wrote:
>> > So we can have portable formatting of uint32_t types. However there is
>> > a catch. Different compilers can use legally subtly different types
>> > though so we need to probe the compiler defined intdef.h first.
>>
>> Interesting, what platform has long uint32_t?  I thought the issue was
>> whether 64-bit is long or "long long".
>>
>> Paolo
>>
>> > Signed-off-by: Alex Bennée <alex.bennee at linaro.org>
>> > ---
>> >  Makefile       |  1 +
>> >  configure      | 13 +++++++++++++
>> >  lib/libcflat.h |  9 +++++++++
>> >  3 files changed, 23 insertions(+)
>> >
>> > diff --git a/Makefile b/Makefile
>> > index a32333b..9822d9a 100644
>> > --- a/Makefile
>> > +++ b/Makefile
>> > @@ -55,6 +55,7 @@ CFLAGS += $(fomit_frame_pointer)
>> >  CFLAGS += $(fno_stack_protector)
>> >  CFLAGS += $(fno_stack_protector_all)
>> >  CFLAGS += $(wno_frame_address)
>> > +CFLAGS += $(if $(U32_LONG_FMT),-D__U32_LONG_FMT__,)
>> >
>> >  CXXFLAGS += $(CFLAGS)
>> >
>> > diff --git a/configure b/configure
>> > index 995c8fa..127868c 100755
>> > --- a/configure
>> > +++ b/configure
>> > @@ -109,6 +109,18 @@ if [ -f $testdir/run ]; then
>> >      ln -fs $testdir/run $testdir-run
>> >  fi
>> >
>> > +# check if uint32_t needs a long format modifier
>> > +cat << EOF > lib_test.c
>> > +#include <inttypes.h>
>> > +EOF
>> > +
>> > +$cross_prefix$cc lib_test.c -E | grep "typedef" | grep "long" | grep "uint32_t" &> /dev/null
>
> This won't work with cross compilers that don't have inttypes.h in their
> path (like mine). How about something like

Hmm good point, in my case inttypes.h came from the libnewlib-dev package.

>
>  cat << EOF > lib_test.c
>  __UINT32_TYPE__
>  EOF
>  u32_long="`aarch64-linux-gnu-gcc lib_test.c -E | awk '/unsigned/ && $2 == "long"'`"

Hmm it is not clear from the docs if this is a GCCism. If it is do we
care? Also the docs do say:

"Some of these macros may not be defined on particular systems if GCC does not provide a stdint.h header on those systems. "

> Although I feel there should be a compiler macro way to do this without
> a need for configure/makefile trickery at all...

I did ask our toolchain bods. They started going on about potential
solutions using _Generic but I fear that might be worse in this case!

>
> Thanks,
> drew
>
>
>> > +exit=$?
>> > +if [ $exit -eq 0 ]; then
>> > +    u32_long=true
>> > +fi
>> > +rm -f lib_test.c
>> > +
>> >  # check for dependent 32 bit libraries
>> >  if [ "$arch" != "arm" ]; then
>> >  cat << EOF > lib_test.c
>> > @@ -155,4 +167,5 @@ TEST_DIR=$testdir
>> >  FIRMWARE=$firmware
>> >  ENDIAN=$endian
>> >  PRETTY_PRINT_STACKS=$pretty_print_stacks
>> > +U32_LONG_FMT=$u32_long
>> >  EOF
>> > diff --git a/lib/libcflat.h b/lib/libcflat.h
>> > index 380395f..e80fc50 100644
>> > --- a/lib/libcflat.h
>> > +++ b/lib/libcflat.h
>> > @@ -58,12 +58,21 @@ typedef _Bool		bool;
>> >  #define true  1
>> >
>> >  #if __SIZEOF_LONG__ == 8
>> > +#  define __PRI32_PREFIX
>> >  #  define __PRI64_PREFIX	"l"
>> >  #  define __PRIPTR_PREFIX	"l"
>> >  #else
>> > +#if defined(__U32_LONG_FMT__)
>> > +#  define __PRI32_PREFIX        "l"
>> > +#else
>> > +#  define __PRI32_PREFIX
>> > +#endif
>> >  #  define __PRI64_PREFIX	"ll"
>> >  #  define __PRIPTR_PREFIX
>> >  #endif
>> > +#define PRId32  __PRI32_PREFIX	"d"
>> > +#define PRIu32  __PRI32_PREFIX	"u"
>> > +#define PRIx32  __PRI32_PREFIX	"x"
>> >  #define PRId64  __PRI64_PREFIX	"d"
>> >  #define PRIu64  __PRI64_PREFIX	"u"
>> >  #define PRIx64  __PRI64_PREFIX	"x"
>> >
>> --
>> To unsubscribe from this list: send the line "unsubscribe kvm" in
>> the body of a message to majordomo at vger.kernel.org
>> More majordomo info at  http://vger.kernel.org/majordomo-info.html


--
Alex Bennée