Remove true/false from GDB ....

Remove true/false from GDB ....

[Andrew Cagney]

Hello,

This is fallout from the recent <stdbool.h> problem.

"bfd.h" was providing ``true'' and ``false'' as convenience 
enums/macros/...  They unfortunatly clash with systems that provide 
<stdbool.h> (a header in c99?) and even some systems that don't.  The 
relevant code block is:

/* I'm sure this is going to break something and someone is going to
    force me to change it.  */
/* typedef enum boolean {false, true} boolean; */
/* Yup, SVR4 has a "typedef enum boolean" in <sys/types.h>  -fnf */
/* It gets worse if the host also defines a true/false enum... -sts */
/* And even worse if your compiler has built-in boolean types... -law */
/* And even worse if your compiler provides a stdbool.h that conflicts
    with these definitions... gcc 2.95 and later do.  If so, it must
    be included first.  -drow */
#if ...
   ... many valiant attemts to define true and false ...
#else
/* Use enum names that will appear nowhere else.  */
typedef enum bfd_boolean {bfd_fffalse, bfd_tttrue} boolean;
#endif

In short, bfd.h should never have been polluting the name space with 
``true'' and ``false''.

So the proposal is for "bfd.h" to remove all the above code and instead 
just define:

   typedef int bfd_boolean;

i.e. 0 is false, non-zero is true, just like C intended :-)

Problem is, some blocks of GDB make use of ``true'' and ``false'' and 
they will need to be changed.  Two possabilities come to mind:

	#include "gdb_stdbool.h"
		which would wrap <stdbool.h>

	zap ``true'' and ``false''

I've strong preferences for the latter.  I think BFD serves as a very 
compelling example of what not to do :-)

thoughts?

I should also note that there is some urgency to this - BFD needs to be 
fixed quickly - preferably before 5.2 of GDB branches.

enjoy,
Andrew

Re: Remove true/false from GDB ....

[Daniel Jacobowitz]

On Fri, Feb 08, 2002 at 06:31:44PM -0500, Andrew Cagney wrote:
> Hello,
> 
> This is fallout from the recent <stdbool.h> problem.
> 
> "bfd.h" was providing ``true'' and ``false'' as convenience 
> enums/macros/...  They unfortunatly clash with systems that provide 
> <stdbool.h> (a header in c99?) and even some systems that don't.  The 
> relevant code block is:
> 
> /* I'm sure this is going to break something and someone is going to
>    force me to change it.  */
> /* typedef enum boolean {false, true} boolean; */
> /* Yup, SVR4 has a "typedef enum boolean" in <sys/types.h>  -fnf */
> /* It gets worse if the host also defines a true/false enum... -sts */
> /* And even worse if your compiler has built-in boolean types... -law */
> /* And even worse if your compiler provides a stdbool.h that conflicts
>    with these definitions... gcc 2.95 and later do.  If so, it must
>    be included first.  -drow */
> #if ...
>   ... many valiant attemts to define true and false ...
> #else
> /* Use enum names that will appear nowhere else.  */
> typedef enum bfd_boolean {bfd_fffalse, bfd_tttrue} boolean;
> #endif
> 
> In short, bfd.h should never have been polluting the name space with 
> ``true'' and ``false''.
> 
> So the proposal is for "bfd.h" to remove all the above code and instead 
> just define:
> 
>   typedef int bfd_boolean;
> 
> i.e. 0 is false, non-zero is true, just like C intended :-)
> 
> Problem is, some blocks of GDB make use of ``true'' and ``false'' and 
> they will need to be changed.  Two possabilities come to mind:
> 
> 	#include "gdb_stdbool.h"
> 		which would wrap <stdbool.h>
> 
> 	zap ``true'' and ``false''
> 
> I've strong preferences for the latter.  I think BFD serves as a very 
> compelling example of what not to do :-)
> 
> thoughts?

Strong preference for the latter here too.  We should not attempt to
use true/false in C.  You never know where they're going to come from -
or not come from.

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

Re: Remove true/false from GDB ....

[Kevin Buettner]

On Feb 8,  6:31pm, Andrew Cagney wrote:

> This is fallout from the recent <stdbool.h> problem.
> 
> "bfd.h" was providing ``true'' and ``false'' as convenience 
> enums/macros/...  They unfortunatly clash with systems that provide 
> <stdbool.h> (a header in c99?) and even some systems that don't.  The 
> relevant code block is:
> 
> /* I'm sure this is going to break something and someone is going to
>     force me to change it.  */
> /* typedef enum boolean {false, true} boolean; */
> /* Yup, SVR4 has a "typedef enum boolean" in <sys/types.h>  -fnf */
> /* It gets worse if the host also defines a true/false enum... -sts */
> /* And even worse if your compiler has built-in boolean types... -law */
> /* And even worse if your compiler provides a stdbool.h that conflicts
>     with these definitions... gcc 2.95 and later do.  If so, it must
>     be included first.  -drow */
> #if ...
>    ... many valiant attemts to define true and false ...
> #else
> /* Use enum names that will appear nowhere else.  */
> typedef enum bfd_boolean {bfd_fffalse, bfd_tttrue} boolean;
> #endif
> 
> In short, bfd.h should never have been polluting the name space with 
> ``true'' and ``false''.
> 
> So the proposal is for "bfd.h" to remove all the above code and instead 
> just define:
> 
>    typedef int bfd_boolean;
> 
> i.e. 0 is false, non-zero is true, just like C intended :-)
> 
> Problem is, some blocks of GDB make use of ``true'' and ``false'' and 
> they will need to be changed.  Two possabilities come to mind:
> 
> 	#include "gdb_stdbool.h"
> 		which would wrap <stdbool.h>
> 
> 	zap ``true'' and ``false''
> 
> I've strong preferences for the latter.  I think BFD serves as a very 
> compelling example of what not to do :-)
> 
> thoughts?

If GDB made widespread use of ``true'' and ``false'', I'd suggest
converting these occurences to ``gdb_true'' and ``gdb_false''.  I've
just looked though and GDB has surprisingly few uses of ``true'' and
``false''.  That being the case, I like Andrew's latter suggestion of
just zapping them.

Here's the results of my search after removing the occurrences of
lines containing true and false in comments:

./memattr.c[34]:   false,			/* hwbreak */
./memattr.c[35]:   false,			/* cache */
./memattr.c[36]:   false				/* verify */
./memattr.c[185]: 	attrib.hwbreak = true;
./memattr.c[187]: 	attrib.hwbreak = false;
./memattr.c[191]: 	attrib.cache = true;
./memattr.c[193]: 	attrib.cache = false;
./memattr.c[197]: 	attrib.verify = true;
./memattr.c[199]: 	attrib.verify = false;
./corelow.c[172]: 	  return (true);
./corelow.c[175]:   return (false);
./irix5-nat.c[437]:   abfd->cacheable = true;
./osfsolib.c[256]:   abfd->cacheable = true;
./solib.c[240]:   abfd->cacheable = true;
./symfile.c[1097]:   sym_bfd->cacheable = true;

Kevin

Re: Remove true/false from GDB ....

[Daniel Jacobowitz]

On Fri, Feb 08, 2002 at 04:54:41PM -0700, Kevin Buettner wrote:
> On Feb 8,  6:31pm, Andrew Cagney wrote:
> 
> > This is fallout from the recent <stdbool.h> problem.
> > 
> > "bfd.h" was providing ``true'' and ``false'' as convenience 
> > enums/macros/...  They unfortunatly clash with systems that provide 
> > <stdbool.h> (a header in c99?) and even some systems that don't.  The 
> > relevant code block is:
> > 
> > /* I'm sure this is going to break something and someone is going to
> >     force me to change it.  */
> > /* typedef enum boolean {false, true} boolean; */
> > /* Yup, SVR4 has a "typedef enum boolean" in <sys/types.h>  -fnf */
> > /* It gets worse if the host also defines a true/false enum... -sts */
> > /* And even worse if your compiler has built-in boolean types... -law */
> > /* And even worse if your compiler provides a stdbool.h that conflicts
> >     with these definitions... gcc 2.95 and later do.  If so, it must
> >     be included first.  -drow */
> > #if ...
> >    ... many valiant attemts to define true and false ...
> > #else
> > /* Use enum names that will appear nowhere else.  */
> > typedef enum bfd_boolean {bfd_fffalse, bfd_tttrue} boolean;
> > #endif
> > 
> > In short, bfd.h should never have been polluting the name space with 
> > ``true'' and ``false''.
> > 
> > So the proposal is for "bfd.h" to remove all the above code and instead 
> > just define:
> > 
> >    typedef int bfd_boolean;
> > 
> > i.e. 0 is false, non-zero is true, just like C intended :-)
> > 
> > Problem is, some blocks of GDB make use of ``true'' and ``false'' and 
> > they will need to be changed.  Two possabilities come to mind:
> > 
> > 	#include "gdb_stdbool.h"
> > 		which would wrap <stdbool.h>
> > 
> > 	zap ``true'' and ``false''
> > 
> > I've strong preferences for the latter.  I think BFD serves as a very 
> > compelling example of what not to do :-)
> > 
> > thoughts?
> 
> If GDB made widespread use of ``true'' and ``false'', I'd suggest
> converting these occurences to ``gdb_true'' and ``gdb_false''.  I've
> just looked though and GDB has surprisingly few uses of ``true'' and
> ``false''.  That being the case, I like Andrew's latter suggestion of
> just zapping them.
> 
> Here's the results of my search after removing the occurrences of
> lines containing true and false in comments:
> 
> ./memattr.c[34]:   false,			/* hwbreak */
> ./memattr.c[35]:   false,			/* cache */
> ./memattr.c[36]:   false				/* verify */
> ./memattr.c[185]: 	attrib.hwbreak = true;
> ./memattr.c[187]: 	attrib.hwbreak = false;
> ./memattr.c[191]: 	attrib.cache = true;
> ./memattr.c[193]: 	attrib.cache = false;
> ./memattr.c[197]: 	attrib.verify = true;
> ./memattr.c[199]: 	attrib.verify = false;
> ./corelow.c[172]: 	  return (true);
> ./corelow.c[175]:   return (false);
> ./irix5-nat.c[437]:   abfd->cacheable = true;
> ./osfsolib.c[256]:   abfd->cacheable = true;
> ./solib.c[240]:   abfd->cacheable = true;
> ./symfile.c[1097]:   sym_bfd->cacheable = true;

So would anyone object if we simply removed all of those?

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

Re: Remove true/false from GDB ....

[Andrew Cagney]

> 
> So would anyone object if we simply removed all of those?

There is something of a policy decision here.  What happens when someone 
submits new code using stdbool.h and ``true'' and ``false''?  Is it 
rejected outright (....),  or as happens now, suggest this may not be 
such a good idea after all.

I guess all that can be done is for the rationale to be clearly documented.

TRUE / FALSE probably fall into the same category.

enjoy,
Andrew

Re: Remove true/false from GDB ....

[Andrew Cagney]

Since there has been no assending view, I'll create a bug report. (my 
todo list is currently overflowing :-()

Andrew