Re: (ITS#3409) libldap_r optimization

[ando@sys-net.it]

Kurt@OpenLDAP.org wrote:

>At 08:33 AM 12/10/2004, raphael.ouazana@linagora.com wrote:
>  
>
>>Selon "ando@sys-net.it" <ando@sys-net.it>:
>>
>>    
>>
>>>The analysis looks correct; a reliable way to detect at configure
>>>time if getaddrinfo is reentrant is needed.
>>>      
>>>
>>Maybe a way to do it is to test if the pointer created by getaddrinfo is
>>different in two different threads. It seems that postgresql uses this
>>method to check that some system calls are reentrant.
>>As I have no "thread-unsafe" getaddrinfo, I can't check if it is a good
>>test or not...
>>    
>>
>
>This is not a good test.  I don't believe there is a good test.
>  
>
Maybe we can just hide the mutexes behind a #ifndef 
IS_GETADDRINFO_THREAD_SAFE, and manually define it where the libc 
documentation declares it to be thread safe.  All in all, it looks like 
something regular users can live without, while aware users should be 
allowed to take advantage of.




    SysNet - via Dossi,8 27100 Pavia Tel: +390382573859 Fax: +390382476497