This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug libstdc++/21295] Configuring g++ library for various locales ?


------- Additional Comments From jpr at essi dot fr  2005-05-01 10:16 -------
Subject: Re:  Configuring g++ library for various locales
 ?

pcarlini at suse dot de wrote:

>------- Additional Comments From pcarlini at suse dot de  2005-05-01 09:06 -------
>  
>
>>That's the point, I do not have these lines (I attach my c++locale.h 
>>from my 4.0.0 by=uild, since it is short).
>>    
>>
>
>Therefore, for some reason, the configury decided not to enable the GNU locale
>model (there are mildly complicated checks for this). Now you can do a few
>things: 1- The most naive one, it trying to rebuild passing *explicitly* 
>--enable-clocale=gnu; 2- Then, you should inspect config.log in the build dir
>and see what's going wrong at configure time.
>
>  
>
Thanks for your help, but I did some search on my own and I found the 
problem, effectively by looking at config.log.

The libstdc++ configure script, at some point (line 5731 and following), 
tries to determine the locale model. For this it builds a test program 
which tries to set the locale to the German one ("de_DE"). In my 
distribution, it is possible to selectively install the locales. Thus I 
installed only the ones which are of interest to me, that is French and 
English. In this case, the test program fails with a SEGV, and the local 
model defaults to generic (which, I assume, means only C and POSIX).

Now I installed also the German locale, and everything seems OK.

Don't you thing that there should be some notice in the doc, telling 
that determining the locale model depends on installing the German 
locale ? Or possibly some documentation about --enable-clocale (to the 
best of my knowledge, it is not documented). Or, much better,  find a 
more accurate way to determine the local model. For instance, at least 
(but I know it won't be sufficient), try the current locale (if any, if 
not cross-compiling, and if different from C/POSIX...) when building C++ 
instead of forcing the German one (which appears to me a little 
Süse-oriented ;-)).

"Der Teufel steckt im Detail"

Best regards

Jean-Paul Rigault





-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=21295


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]