Re: [patch] localization testing

[François Dumont <francois dot cppdevs at free dot fr>]

On 09/12/2010 10:30 PM, Paolo Carlini wrote:
> On 09/12/2010 10:08 PM, François Dumont wrote:
>    
> > Hi
> >
> >        I noticed that all the localization tests were considered as
> > unsupported on my platform despite I do have localization support. The
> > problem was in the dg-require-namedlocale deja gnu extension. This
> > proc consider that there is no localization support if any of the 24
> > tested locales is missing and I have one missing. This patch change
> > the behavior of this proc, now it takes the name of the locale to
> > check  and only test this one. I have updated all tests to check for
> > the required locale(s).
> >      
> Heroic! Seriously, quite a bit of rather boring work, but the patch
> indeed makes sense and it's an improvement. Personally, I was under the
> impression that with many distros it's normally an all or nothing thing
> (for instance was always the case for my personal machines), but
> certainly isn't necessarily true.
>
> Now, since you did all this work, I think we should definitely update
> the documentation to reflect the improvement: can you also patch
> doc/xml/manual/prerequisites.xml?
>
>    
The doc is not so bad already, it depends on how you interpret the 
"because of this, the libstdc++ testsuite will skip the named local 
tests". What about simply adding: "the libstdc++ testsuite will skip the 
named locale tests that are missing the "C" information". ? If it is ok, 
do you want me to resubmit the whole patch in addition to the doc part ?

> >        Tested under x86_64-linux. Some of the tests that was
> > unsupported are now failing on my platform. Looks like  I am missing
> > the libstdc++ catalog saying that please stands for 'bitte' in german
> > and 's'il vous plaît' in french. I have also a problem with a monetary
> > format in money_put/put/char/9780-3.cc, with glibc 2.11.1 I should
> > have '-12.345.678,90 EUR ' but I have '-EUR  12.345.678,90'. Nothing
> > that makes this patch invalid.
> >      
> The last issue is weird, doesn't happen with 2.11.2 here, and apparently
> nobody noticed anything so far. Can you investigate a bit further
> independently of the present patch? We could also allow both forms, we
> did something similar in the past...
>    
Ok I will try to understand the money format issue. What about the 
other, am I missing something, a libstdc++ catalog ?

Francois