Confusion over the definition of 'bool' in rdi-share/host.h

Confusion over the definition of 'bool' in rdi-share/host.h

[Nick Clifton]

Hi Guys,

  The ARM port of GDB is currently failing to build for me because of
  this problem:

In file included from /home/nickc/work/sources/egcs/gdb/rdi-share/ardi.h:17,
                 from /home/nickc/work/sources/egcs/gdb/remote-rdi.c:46:
/home/nickc/work/sources/egcs/gdb/rdi-share/host.h:123: conflicting types for `_Bool'
/usr/lib/gcc-lib/i386-redhat-linux/2.96/include/stdbool.h:41: previous declaration of `_Bool'

  It seems that host.h has code that looks like this (trimmed a little):

    #  define _bool int

    #ifdef _bool
       typedef _bool bool;
    #endif

  And stdbool.h has:

    typedef enum
    {
      false = 0,
      true = 1
    } _Bool;

    #define bool _Bool

  So the typedef in host.h becomes, effectively  "typedef int enum _Bool".

  I am not sure if we are allowed to modify rdi-share/host.h, it
  appears to be copyright to ARM, but if we are, then may I submit the
  following patch to undefine bool before it is used ?

Cheers
        Nick

2002-02-05  Nick Clifton  <nickc@cambridge.redhat.com>

	* rdi-share/host.h: Before creating a typedef from _bool, make
        sure that bool is not defined.

Index: gdb/rdi-share/host.h
===================================================================
RCS file: /cvs/src/src/gdb/rdi-share/host.h,v
retrieving revision 1.4
diff -c -3 -p -w -r1.4 host.h
*** host.h	2001/06/29 01:19:23	1.4
--- host.h	2002/02/05 17:48:54
*************** typedef unsigned       char  unsigned8;
*** 120,125 ****
--- 120,128 ----
  #  if defined(_MFC_VER) || defined(__CC_NORCROFT) /* When using MS Visual C/C++ v4.2 */
  #    define bool _bool /* avoids "'bool' is reserved word" warning      */
  #  else
+ #    ifdef bool
+ #    undef bool
+ #    endif
       typedef _bool bool;
  #  endif
  #  define true _true

Re: Confusion over the definition of 'bool' in rdi-share/host.h

[Daniel Jacobowitz]

On Tue, Feb 05, 2002 at 06:12:16PM +0000, Nick Clifton wrote:
> Hi Guys,
> 
>   The ARM port of GDB is currently failing to build for me because of
>   this problem:
> 
> In file included from /home/nickc/work/sources/egcs/gdb/rdi-share/ardi.h:17,
>                  from /home/nickc/work/sources/egcs/gdb/remote-rdi.c:46:
> /home/nickc/work/sources/egcs/gdb/rdi-share/host.h:123: conflicting types for `_Bool'
> /usr/lib/gcc-lib/i386-redhat-linux/2.96/include/stdbool.h:41: previous declaration of `_Bool'
> 
>   It seems that host.h has code that looks like this (trimmed a little):
> 
>     #  define _bool int
> 
>     #ifdef _bool
>        typedef _bool bool;
>     #endif
> 
>   And stdbool.h has:
> 
>     typedef enum
>     {
>       false = 0,
>       true = 1
>     } _Bool;
> 
>     #define bool _Bool
> 
>   So the typedef in host.h becomes, effectively  "typedef int enum _Bool".
> 
>   I am not sure if we are allowed to modify rdi-share/host.h, it
>   appears to be copyright to ARM, but if we are, then may I submit the
>   following patch to undefine bool before it is used ?

(Ew on all counts)

Does not defining bool if it is already defined also work?

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: Confusion over the definition of 'bool' in rdi-share/host.h

[Stan Shebs]

Nick Clifton wrote:
> 
>   I am not sure if we are allowed to modify rdi-share/host.h, it
>   appears to be copyright to ARM, but if we are, then may I submit the
>   following patch to undefine bool before it is used ?

We do whatever we want to rdi-share/*.  My personal vote is to
delete it and start over... :-)

Stan

Re: Confusion over the definition of 'bool' in rdi-share/host.h

[Andrew Cagney]

> Nick Clifton wrote:
> 
>> 
>> I am not sure if we are allowed to modify rdi-share/host.h, it
>> appears to be copyright to ARM, but if we are, then may I submit the
>> following patch to undefine bool before it is used ?
> 
> 
> We do whatever we want to rdi-share/*.  My personal vote is to
> delete it and start over... :-)


Next release? :-) Have you seen how much protocol/remote stuff deleting 
the a29k eliminates?

Andrew

Re: Confusion over the definition of 'bool' in rdi-share/host.h

[Grant Edwards]

On Tue, Feb 05, 2002 at 10:29:00PM -0500, Andrew Cagney wrote:
> > Nick Clifton wrote:
> > 
> >> 
> >> I am not sure if we are allowed to modify rdi-share/host.h, it
> >> appears to be copyright to ARM, but if we are, then may I submit the
> >> following patch to undefine bool before it is used ?
> > 
> > 
> > We do whatever we want to rdi-share/*.  My personal vote is to
> > delete it and start over... :-)
>
> Next release? :-) Have you seen how much protocol/remote stuff
> deleting the a29k eliminates?

OK, I see the smileys, but this would be a major problem for
me, so forgive my paranoia:

Are you seriously proposing that somebody re-write ADP protocol
support from scratch or that support for ADP be abandoned
completely?

Somebody at EPI is working on supporting the "remote" protocol
with newer JTAG interfaces, but until that happens (and
somebody gives me about $10K to replace my development
hardware), I need ADP support.

-- 
Grant Edwards
grante@visi.com

Re: Confusion over the definition of 'bool' in rdi-share/host.h

[Stan Shebs]

Grant Edwards wrote:
> 
> Are you seriously proposing that somebody re-write ADP protocol
> support from scratch or that support for ADP be abandoned
> completely?

Speaking seriously, ADP is still in use even if it is a lame
protocol, so no it's not going to get dropped.  The RDI code is
however a completely steamy pile, and if I had to do it all over
again, I would rather have spent three weeks of 16-hour days
implementing the protocol from scratch, rather than taking
the same amount of time stretched out over months trying to
get it to work right on every host.

The ADP protocol is more complicated than the standard
remote protocol, and if you stick to what actually works rather
than what ARM fantasized about but never implemented, it's
not really that hairy.  Another case where I chickened out
rather than doing the Right Thing.

Stan

Re: Confusion over the definition of 'bool' in rdi-share/host.h

[Grant Edwards]

On Wed, Feb 06, 2002 at 09:36:02PM -0800, Stan Shebs wrote:
> 
> Speaking seriously, ADP is still in use even if it is a lame
> protocol, so no it's not going to get dropped.  The RDI code is
> however a completely steamy pile, and if I had to do it all over
> again, I would rather have spent three weeks of 16-hour days
> implementing the protocol from scratch, rather than taking
> the same amount of time stretched out over months trying to
> get it to work right on every host.

Yup.  I've spent god-knows how many hours fussing with the code
that parses device names because it keeps refusing to recognize
serial ports on my machine.  Apparently that bit of code was
some sort of failed AI experiment.  You'd think that if I claim
I've got a serial port called /dev/ttySI15, then it would just
go ahead and open it!  But noooo, the RDI code thinks it has a
priori knowledge about what serial ports I have and what they
are called.

> The ADP protocol is more complicated than the standard remote
> protocol,

That's the understatement of the week!  :)

ADP has more unused features than anything I've seen in a
_long_ time.  It looks more like a PhD project than something
somebody actually expected to work with.

> and if you stick to what actually works rather than what ARM
> fantasized about but never implemented, it's not really that
> hairy.  Another case where I chickened out rather than doing
> the Right Thing.

It happens to all of us...

-- 
Grant Edwards
grante@visi.com

Re: Confusion over the definition of 'bool' in rdi-share/host.h

[Richard Earnshaw]

> Hi Guys,
> 
>   The ARM port of GDB is currently failing to build for me because of
>   this problem:
> 
> In file included from /home/nickc/work/sources/egcs/gdb/rdi-share/ardi.h:17,
>                  from /home/nickc/work/sources/egcs/gdb/remote-rdi.c:46:
> /home/nickc/work/sources/egcs/gdb/rdi-share/host.h:123: conflicting types for `_Bool'
> /usr/lib/gcc-lib/i386-redhat-linux/2.96/include/stdbool.h:41: previous declaration of `_Bool'
> 
>   It seems that host.h has code that looks like this (trimmed a little):
> 
>     #  define _bool int
> 
>     #ifdef _bool
>        typedef _bool bool;
>     #endif
> 
>   And stdbool.h has:
> 
>     typedef enum
>     {
>       false = 0,
>       true = 1
>     } _Bool;
> 
>     #define bool _Bool
> 
>   So the typedef in host.h becomes, effectively  "typedef int enum _Bool".
> 
>   I am not sure if we are allowed to modify rdi-share/host.h, it
>   appears to be copyright to ARM, but if we are, then may I submit the
>   following patch to undefine bool before it is used ?
> 
> Cheers
>         Nick

the rdi's host.h is really a non-autoconf approach to generating host 
definitions.  It's contents should really be replaced with an include of 
the standard autoconf generated definitions followed by some cleanups (ie 
defining the types that the rdi code need in terms of what autoconf has 
provided) to make the rest of the rdi files compile cleanly.

R.

Re: Confusion over the definition of 'bool' in rdi-share/host.h

[Fernando Nasser]

Richard Earnshaw wrote:
> 
> > Hi Guys,
> >
> >   The ARM port of GDB is currently failing to build for me because of
> >   this problem:
> >
> > In file included from /home/nickc/work/sources/egcs/gdb/rdi-share/ardi.h:17,
> >                  from /home/nickc/work/sources/egcs/gdb/remote-rdi.c:46:
> > /home/nickc/work/sources/egcs/gdb/rdi-share/host.h:123: conflicting types for `_Bool'
> > /usr/lib/gcc-lib/i386-redhat-linux/2.96/include/stdbool.h:41: previous declaration of `_Bool'
> >
> >   It seems that host.h has code that looks like this (trimmed a little):
> >
> >     #  define _bool int
> >
> >     #ifdef _bool
> >        typedef _bool bool;
> >     #endif
> >
> >   And stdbool.h has:
> >
> >     typedef enum
> >     {
> >       false = 0,
> >       true = 1
> >     } _Bool;
> >
> >     #define bool _Bool
> >
> >   So the typedef in host.h becomes, effectively  "typedef int enum _Bool".
> >
> >   I am not sure if we are allowed to modify rdi-share/host.h, it
> >   appears to be copyright to ARM, but if we are, then may I submit the
> >   following patch to undefine bool before it is used ?
> >
> > Cheers
> >         Nick
> 
> the rdi's host.h is really a non-autoconf approach to generating host
> definitions.  It's contents should really be replaced with an include of
> the standard autoconf generated definitions followed by some cleanups (ie
> defining the types that the rdi code need in terms of what autoconf has
> provided) to make the rest of the rdi files compile cleanly.
> 
> R.

Yes.

But until someone finds the time to do that I guess we must apply
Nick's patch.

Regards to all,
Fernando


-- 
Fernando Nasser
Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: Confusion over the definition of 'bool' in rdi-share/host.h

[Grant Edwards]

On Wed, Feb 06, 2002 at 09:12:28AM -0500, Fernando Nasser wrote:

> > the rdi's host.h is really a non-autoconf approach to
> > generating host definitions.  It's contents should really be
> > replaced with an include of the standard autoconf generated
> > definitions followed by some cleanups (ie defining the types
> > that the rdi code need in terms of what autoconf has provided)
> > to make the rest of the rdi files compile cleanly.

> But until someone finds the time to do that I guess we must
> apply Nick's patch.

I could probably make some time to work on this if somebody
wanted to tell me exactly what set of sources I should start
with.  Keeping support for ADP is pretty important to one of
the products I work on.

-- 
Grant Edwards
grante@visi.com