warning compiling dbusbind.c

warning compiling dbusbind.c

[Glenn Morris]

I receive this warning compiling dbusbind.c on x86_64 GNU/Linux with
gcc 4.1:

dbusbind.c: In function 'xd_retrieve_arg':
dbusbind.c:600: warning: comparison is always false due to limited
range of data type

Re: warning compiling dbusbind.c

[Michael Albinus]

Glenn Morris <rgm@gnu.org> writes:

> I receive this warning compiling dbusbind.c on x86_64 GNU/Linux with
> gcc 4.1:
>
> dbusbind.c: In function 'xd_retrieve_arg':
> dbusbind.c:600: warning: comparison is always false due to limited
> range of data type

The code in question is

	dbus_uint32_t val;
	dbus_message_iter_get_basic (iter, &val);
	XD_DEBUG_MESSAGE ("%c %d", dtype, val);
	return make_fixnum_or_float (val);

I've used make_fixnum_or_float, because BITS_PER_EMACS_INT is too small
on 32bit machines. One possible solution could be

	dbus_uint32_t val;
	dbus_message_iter_get_basic (iter, &val);
	XD_DEBUG_MESSAGE ("%c %d", dtype, val);
#if BITS_PER_EMACS_INT >= 32
	return make_number (val);
#else
	return make_fixnum_or_float (val);
#endif

On the other hand, shouldn't it be handled in make_fixnum_or_float?

Best regards, Michael.

Re: warning compiling dbusbind.c

[Stefan Monnier]

> I've used make_fixnum_or_float, because BITS_PER_EMACS_INT is too small
> on 32bit machines. One possible solution could be

> 	dbus_uint32_t val;
> 	dbus_message_iter_get_basic (iter, &val);
> 	XD_DEBUG_MESSAGE ("%c %d", dtype, val);
> #if BITS_PER_EMACS_INT >= 32
> 	return make_number (val);
> #else
> 	return make_fixnum_or_float (val);
> #endif

Yuck.  I must prefer living with the warning (which is just that:
a warning.  We already have several similar ones).

> On the other hand, shouldn't it be handled in make_fixnum_or_float?

No, because make_fixnum_or_float expects an integer of type EMACS_INT.
The problem here is that you pass it an object that is slight smaller,
so soome of its internal code cann be optimized away, which is what the
compiler warns you about.  Maybe casting val to EMACS_INT is enough to
silence the compiler?

I wonder if gcc would also complain  in the case where
make_fixnum_or_float is an inlinable function.  I'd guess not.


        Stefan

Re: warning compiling dbusbind.c

[Glenn Morris]

Stefan Monnier wrote:

> Yuck.  I must prefer living with the warning (which is just that:
> a warning.  We already have several similar ones).

I menioned this because it is the only warning of any kind I get when
compiling the Emacs C code.

(There are a few instances of workarounds for this in the C code
already: `grep whining src/*.c' .)

Re: warning compiling dbusbind.c

[Nick Roberts]

 > > Yuck.  I must prefer living with the warning (which is just that:
 > > a warning.  We already have several similar ones).
 > 
 > I menioned this because it is the only warning of any kind I get when
 > compiling the Emacs C code.

Indeed, in GDB all files are compiled with the -Werror option, presumably
because GCC folks are more knowledgeable about what is undesirable than the
average developer.

-- 
Nick                                           http://www.inet.net.nz/~nickrob

Re: warning compiling dbusbind.c

[Stefan Monnier]

>> > Yuck.  I must prefer living with the warning (which is just that:
>> > a warning.  We already have several similar ones).
>> 
>> I menioned this because it is the only warning of any kind I get when
>> compiling the Emacs C code.

> Indeed, in GDB all files are compiled with the -Werror option, presumably
> because GCC folks are more knowledgeable about what is undesirable than the
> average developer.

But uglifying the code and making it less maintainable just to get rid
of a minor warning is not good software engineering practices.

I'd much rather introduce a dummy macro like

    #define DEFEAT_WARNING(x) ((EMACS_INT)(x))

or even

    #define DEFEAT_WARNING(x) ((((EMACS_INT)(x))+1)-1)

then introduce the "#if BITS_PER_EMACS_INT >= 32" that dispatches to two
different macros, where the reason why it's correct is not nearly
as obvious.


        Stefan

Re: warning compiling dbusbind.c

[Michael Albinus]

Glenn Morris <rgm@gnu.org> writes:

> I menioned this because it is the only warning of any kind I get when
> compiling the Emacs C code.
>
> (There are a few instances of workarounds for this in the C code
> already: `grep whining src/*.c' .)

Thanks for the pointer, I've fixed it this way. When there is a macro,
as Stefan has proposed, it could be introduced for all these cases at
once.

Could you, please, check on your machine? I haven't a 64bit one.

Best regards, Michael.

Re: warning compiling dbusbind.c

[Glenn Morris]

Michael Albinus wrote:

> Thanks for the pointer, I've fixed it this way. When there is a macro,
> as Stefan has proposed, it could be introduced for all these cases at
> once.
>
> Could you, please, check on your machine? I haven't a 64bit one.

No more compilation warning; thanks.