Warning when using qint64 with PRIi64 on 64 bit linux

[DuBu]

Hi,

when I try to print a qint64 with printf and the format specifier PRIi64 on a 64 bit linux I get the following warning:

warning: format ‘%li’ expects argument of type ‘long int’, but argument 4 has type ‘qint64 {aka long long int}’ [-Wformat=]                                                                                               

g++ --version:
g++ (Ubuntu 5.4.0-6ubuntu1~16.04.5) 5.4.0 20160609

How can I print without warnings?

Thanks in advance!

[m.sue]

Hi @DuBu

PRIi64 should return "lli" on a 64-bit system. So maybe you use a 32-bit system or (cross-)compiler.

-Michael.

[jsulm]

@DuBu Use ll instead of li, see http://www.cplusplus.com/reference/cstdio/printf/

[DuBu]

@m.sue Yes, I thought so too. But it doesn't! Here's a snippet from inttypes.h:

# if __WORDSIZE == 64
#  define __PRI64_PREFIX	"l"
#  define __PRIPTR_PREFIX	"l"
# else
#  define __PRI64_PREFIX	"ll"
#  define __PRIPTR_PREFIX
# endif
# define PRIi64		__PRI64_PREFIX "i"

And the if branch is active (i.e __WORDSIZE == 64 is true).
So PRIi64 returns "li" on my system. The compiler is the one I already mentioned.

[m.sue]

Hi @DuBu

I have the same compiler and the same code as you in /usr/include/inttypes.h. But PRIi64 returns "lli". No idea what could be the difference :-/

Ah I just tested: I get the same error as you in main.cpp but in some sub-dll where I use all my PRI* strings everything works fine. Interesting.

-Michael.

[m.sue]

Hi @DuBu

It works if you use int64_t instead of qint64.

-Michael.

[DuBu]

@m.sue I can't change that, it's code from Qt (i.e QElapsedTimer::elapsed()).

[m.sue]

Hi @DuBu

You could probably cast it to int64_t. It is ugly, though.

-Michael.

[DuBu]

@jsulm Yes, but I want to use the PRI macros.

[DuBu]

I wonder what is wrong, the qint64 which is actually a long long int or the PRIi64 macro which is actually a "li"?

[m.sue]

Hi @DuBu

PRIi64 together with int64_t is "lli", together with qint64 it is "li". The PRI macros probably do not know what to do about qint64. Maybe it's a compiler bug.

-Michael.

[DuBu]

Ok, thanks guys! I ended up by using PRIi64 and casting qint64 to int64_t when necessary.

[m.sue]

Hi @DuBu

Qt could fix it by defining typedef int64_t qint64; instead of typedef long long qint64; on Linux. But, of course, there could be some side effects of that change that I do not see. Maybe some supported compiler does not yet have the int64_t type.

-Michael.

[jsulm]

@m.sue long long is not necessarily always 64 bit. The length of integral types in C/C++ isn't specified exactly. The only thing specified is:

char <= short <= int <= long int <= long long

The exact size of each type depends on platform and compiler.

[m.sue]

Hi @jsulm

Ok, but as int64_t is guarantueed to be 64-bit, it is not guarantueed to be defined for every compiler. So it's no good choice for qint64.

Nevertheless Qt should find a way to use it without warnings in printf with the new PRI macros, They are just about invented to use with long longs that are or are not 64-bit depending on the compiler/machine.

-Michael.

[jsulm]

@m.sue Well, qint64 is guaranteed to be 64 bit, so you cannot use long long to typedef it.
int64_t is defined in C++11 and as Qt requires C++11 since quite some time it is perfectly valid to use int64_t in my opinion.
And why use printf in C++ at all?

[DuBu]

@jsulm Yes, to use int64_t is perfectly valid. So Qt should use it in their own classes.
Why use printf in C++? Cause it sits deep down in a C library we want to use in C++ as well. And also: why not?