Broken dbusbind.c

Broken dbusbind.c

[John Yates]

I am trying to build the last tip.  I get compile errors in dbusbind.c.
The culprit is a broken macro definition:

/* Check whether TYPE is a basic DBusType.  */
#ifdef HAVE_DBUS_TYPE_IS_VALID
#define XD_BASIC_DBUS_TYPE(type) \
  (dbus_type_is_valid (type) && dbus_type_is_basic (type))
#else
#define XD_BASIC_DBUS_TYPE(type) \  <<<< definition must be a single
logical line
  ((type == DBUS_TYPE_BYTE) \
   || (type == DBUS_TYPE_BOOLEAN) \
   || (type == DBUS_TYPE_INT16) \
   || (type == DBUS_TYPE_UINT16) \
   || (type == DBUS_TYPE_INT32) \
   || (type == DBUS_TYPE_UINT32) \
   || (type == DBUS_TYPE_INT64) \
   || (type == DBUS_TYPE_UINT64) \
   || (type == DBUS_TYPE_DOUBLE) \
   || (type == DBUS_TYPE_STRING) \
   || (type == DBUS_TYPE_OBJECT_PATH) \
   || (type == DBUS_TYPE_SIGNATURE) \
#ifdef DBUS_TYPE_UNIX_FD             <<<< Oops!
   || (type == DBUS_TYPE_UNIX_FD) \
#endif
   )
#endif

Re: Broken dbusbind.c

[Lars Ingebrigtsen]

John Yates <john@yates-sheets.org> writes:

> I am trying to build the last tip.  I get compile errors in dbusbind.c.
> The culprit is a broken macro definition:

You don't specify the OS this breaks on -- the current Emacs trunk
builds fine for me on Debian bullseye, for instance.

-- 
(domestic pets only, the antidote for overdose, milk.)
   bloggy blog: http://lars.ingebrigtsen.no

Re: Broken dbusbind.c

[John Yates]

On Mon, Dec 28, 2020 at 7:40 PM Lars Ingebrigtsen <larsi@gnus.org> wrote:
>
> You don't specify the OS this breaks on -- the current Emacs trunk
> builds fine for me on Debian bullseye, for instance.

Here is more context:

/* Check whether TYPE is a basic DBusType.  */
#ifdef HAVE_DBUS_TYPE_IS_VALID
#define XD_BASIC_DBUS_TYPE(type) \
  (dbus_type_is_valid (type) && dbus_type_is_basic (type))
#else
#define XD_BASIC_DBUS_TYPE(type) \
  ((type == DBUS_TYPE_BYTE) \
   || (type == DBUS_TYPE_BOOLEAN) \
   || (type == DBUS_TYPE_INT16) \
   || (type == DBUS_TYPE_UINT16) \
   || (type == DBUS_TYPE_INT32) \
   || (type == DBUS_TYPE_UINT32) \
   || (type == DBUS_TYPE_INT64) \
   || (type == DBUS_TYPE_UINT64) \
   || (type == DBUS_TYPE_DOUBLE) \
   || (type == DBUS_TYPE_STRING) \
   || (type == DBUS_TYPE_OBJECT_PATH) \
   || (type == DBUS_TYPE_SIGNATURE) \
#ifdef DBUS_TYPE_UNIX_FD
   || (type == DBUS_TYPE_UNIX_FD) \
#endif
   )
#endif

Notice that, if HAVE_DBUS_TYPE_IS_VALID is defined
we get the first #define XD_BASIC_DBUS_TYPE and
all will be well.

OTOH, if HAVE_DBUS_TYPE_IS_VALID is undefined
we get the broken #define.  (Imagine what the C lexer
sees when each \<NL> disappears.  The result is a line
with a #ifdef DBUS_TYPE_UNIX_FD tacked onto the
end.)

My sense is that the OS is immaterial.  Is not the
HAVE_DBUS_TYPE_IS_VALID symbol supplied by
the configure script?  Hence it comes down to what
headers and libraries have been installed.

(For what it is worth I am on Ubuntu 20.10.)

/john

Re: Broken dbusbind.c

[Michael Albinus]

John Yates <john@yates-sheets.org> writes:

Hi John,

> Here is more context:
>
> /* Check whether TYPE is a basic DBusType.  */
> #ifdef HAVE_DBUS_TYPE_IS_VALID
> #define XD_BASIC_DBUS_TYPE(type) \
>   (dbus_type_is_valid (type) && dbus_type_is_basic (type))
> #else
> #define XD_BASIC_DBUS_TYPE(type) \
>   ((type == DBUS_TYPE_BYTE) \
>    || (type == DBUS_TYPE_BOOLEAN) \
>    || (type == DBUS_TYPE_INT16) \
>    || (type == DBUS_TYPE_UINT16) \
>    || (type == DBUS_TYPE_INT32) \
>    || (type == DBUS_TYPE_UINT32) \
>    || (type == DBUS_TYPE_INT64) \
>    || (type == DBUS_TYPE_UINT64) \
>    || (type == DBUS_TYPE_DOUBLE) \
>    || (type == DBUS_TYPE_STRING) \
>    || (type == DBUS_TYPE_OBJECT_PATH) \
>    || (type == DBUS_TYPE_SIGNATURE) \
> #ifdef DBUS_TYPE_UNIX_FD
>    || (type == DBUS_TYPE_UNIX_FD) \
> #endif
>    )
> #endif
>
> Notice that, if HAVE_DBUS_TYPE_IS_VALID is defined
> we get the first #define XD_BASIC_DBUS_TYPE and
> all will be well.
>
> OTOH, if HAVE_DBUS_TYPE_IS_VALID is undefined
> we get the broken #define.  (Imagine what the C lexer
> sees when each \<NL> disappears.  The result is a line
> with a #ifdef DBUS_TYPE_UNIX_FD tacked onto the
> end.)

Might be. However, ...

> (For what it is worth I am on Ubuntu 20.10.)

... I'm also testing on Ubuntu 20.10. Why is HAVE_DBUS_TYPE_IS_VALID
undefined for you?

> /john

Best regards, Michael.

Re: Broken dbusbind.c

[Michael Albinus]

Michael Albinus <michael.albinus@gmx.de> writes:

Hi John,

>> OTOH, if HAVE_DBUS_TYPE_IS_VALID is undefined
>> we get the broken #define.  (Imagine what the C lexer
>> sees when each \<NL> disappears.  The result is a line
>> with a #ifdef DBUS_TYPE_UNIX_FD tacked onto the
>> end.)

I've fixed this in master.

>> (For what it is worth I am on Ubuntu 20.10.)
>
> ... I'm also testing on Ubuntu 20.10. Why is HAVE_DBUS_TYPE_IS_VALID
> undefined for you?

This is still not clear to me; I'm curious about.

>> /john

Best regards, Michael.

Re: Broken dbusbind.c

[John Yates]

On Tue, Dec 29, 2020 at 3:21 PM Michael Albinus <michael.albinus@gmx.de> wrote:
>
> Why is HAVE_DBUS_TYPE_IS_VALID
> undefined for you?

I do not know.  I have a random collection of -dev packages
installed which I have been determined empirically over time
be needed to build the projects I track.  I do not know what
feature configure tests for when deciding whether or not to
define HAVE_DBUS_TYPE_IS_VALID.  Do you?

That said, the #else clause is clearly broken.  Do you agree?
If not then I challenge you do force its use and see that the
result does not compile.

/john

Re: Broken dbusbind.c

[Michael Albinus]

John Yates <john@yates-sheets.org> writes:

Hi John,

>> Why is HAVE_DBUS_TYPE_IS_VALID
>> undefined for you?
>
> I do not know.  I have a random collection of -dev packages
> installed which I have been determined empirically over time
> be needed to build the projects I track.  I do not know what
> feature configure tests for when deciding whether or not to
> define HAVE_DBUS_TYPE_IS_VALID.  Do you?

# pkg-config --cflags dbus-1
-I/usr/include/dbus-1.0 -I/usr/lib/x86_64-linux-gnu/dbus-1.0/include

> That said, the #else clause is clearly broken.  Do you agree?
> If not then I challenge you do force its use and see that the
> result does not compile.

You might have seen that I did it already yesterday, as said in the
other message.

> /john

Best regards, Michael.

Re: Broken dbusbind.c

[John Yates]

On Thu, Dec 31, 2020 at 4:00 AM Michael Albinus <michael.albinus@gmx.de> wrote:
>
> You might have seen that I did it already yesterday, as said in the
> other message.

Missed that.  My bad.  Thank you.

/john