[Python-Dev] Dealing with test__locale failure on OS X before a3

[Ronald Oussoren]

On 27-aug-04, at 2:49, Nick Bastin wrote:

>
> On Aug 26, 2004, at 7:41 PM, Brett C. wrote:
>
>> Martin v. Löwis wrote:
>>
>>> Brett C. wrote:
>>>> Personally I would like to turn off the test on OS X for now and 
>>>> then try to see if there is a proper way to solve this.
>>> In what way does that help? The test shows that the locale module is
>>> broken on OS X. So the only sensible options are to remove the module
>>> (or atleast the relevant functions), or to fix them.
>>
>> I have an idea for detecting a broken setlocale.  When configure.in 
>> does its compiling does it compile with the same flags that the 
>> Makefile uses for compiling?  If it does we can test for something we 
>> know will be different (fr_FR has a different decimal point, for 
>> instance) and compare them and then have a value in pyconfig.h to 
>> signal that a busted setlocale exists.  setlocale can then raise 
>> locale.Error automatically for platforms that have that macro 
>> defined.  But the test has to be compiled with the ``-framework 
>> CoreFoundation`` option or else this won't work.
>
> I think we should actually just fix this, rather than removing locale 
> support for MacOS X builds.  The trick is that we need to call 
> __setonlyClocaleconv() again, to remove the protection against calling 
> setlocale() in framework builds (why they did this in the first place 
> is beyond me).  Seems like it should be an easy gate against darwin, 
> and then call that on startup.
Why do you think it is save to re-enable setlocale? There must be some 
reason for disabling setlocale. If you re-enable setlocale some random 
CoreFoundation function might suddenly fail.

Ronald