Re: reasons for having no inclusion guards in config.h

[John Calcote]

On 3/7/2010 2:40 PM, Tim Rice wrote:
> On Sun, 7 Mar 2010, John Calcote wrote:
>
>    
> > On 3/7/2010 11:53 AM, Bob Friesenhahn wrote:
> >      
> [...]
>    
> > > There may be several config.h files, each of which has somewhat different
> > > content.  Bing!
> > >        
> > So what would be wrong with generating an include guard tag based on the name
> > of the project?
> >
> > For instance,
> >
> > --------
> > #ifndef GRAPHICSMAGICK_CONFIG_H
> > #define GRAPHICSMAGICK_CONFIG_H
> >
> > ...
> >
> > #endif
> > --------
> >
> > John
> >      
> To give a real world example, in freeradius there are
> src/modules/rlm_counter/config.h.in
> src/modules/rlm_eap/types/rlm_eap_peap/config.h.in
> src/modules/rlm_eap/types/rlm_eap_tls/config.h.in
> src/modules/rlm_eap/types/rlm_eap_ttls/config.h.in
> src/modules/rlm_example/config.h.in
> src/modules/rlm_ippool/config.h.in
> src/modules/rlm_pam/config.h.in
> src/modules/rlm_perl/config.h.in
> src/modules/rlm_radutmp/config.h.in
> src/modules/rlm_smb/config.h.in
> src/modules/rlm_smsotp/config.h.in
> src/modules/rlm_sql/drivers/rlm_sql_mysql/config.h.in
> src/modules/rlm_sqlhpwippool/config.h.in
> src/modules/rlm_unix/config.h.in
>
> It would work to have each of those with
> #ifndef FREERADIUS_CONFIG_H
> #define FREERADIUS_CONFIG_H
> ....
> #endif
>    

Clearly not. I wasn't considering the case where there were multiple 
config.h files in the *same* project. I assumed Bob was talking about 
including config.h files from different projects. Now I see that that 
would be a problem.

Regardless, there are ways around this issue. All config.h files can't 
be in the same directory, even if in the same project. So a path-based 
include-guard symbol could be generated, e.g.,

#ifndef SRC_MODULES_RLM_COUNTER_CONFIG_H
...

The idea here of course is that the unwanted situation whereby the same 
config.h file is inadvertently included in the same translation unit 
twice could be avoided.

As an aside: It's interesting to note that gcc's preprocessor has 
built-in functionality to automatically guard against this situation.

John