Re: Building on Mac OS/X 10.3

Hi all,

I've just downloaded and compiled the latest release (0.9.7) on OSX 
10.3. I've seen the same problems and found that the "getLogger()" and 
"getRootLogger()" calls cause segfaults unless I call

BasicConfigurator::configure();

first. So, for example delayedloop can be fixed like this:

int main(int argc, char **argv) {
  BasicConfigurator::configure();
  DelayedLoop::logger = Logger::getLogger(_T("DelayedLoop"));
  ...
}

(with logger made public).

This is a bit awkward, because with this it is not possible to assign a 
static logger globally. I don't have another platform available for 
testing but I assume that calling "getLogger()" on other platforms 
works even without having configured first.

I haven't investigated the cause of the crash any further. I recently 
had a problem with code that ran on Linux and crashed on OSX where the 
cause was an uninitialized variable. If I remember correctly, the 
variable was automatically assigned 0 on Linux and uninitialized use 
was 'safe' there while it was containing garbage on OSX and cause a 
crash. As initialization seems to be the problem here, perhaps this 
might give a lead on what's wrong.

HTH,
Sven

Michaël CATANZARITI wrote:

> Hello Curt,
>
> The serversocket.cpp file was fixed in the CVS repository in May.
> I had many problems running the delayedloop example, without finding a 
> solution.
> If you can help ...
>
> Regards,
>
> Curt Arnold wrote:
>
> > > I ran into a few minor gotchas while trying to build log4cxx-0.9.7 
> > on > Mac OS/X. Since nobody else has reported their Mac build 
> > experiences, I > thought I pass along mine. > > autogen.sh calls 
> > libtoolize which is appears to be called glibtoolize on > Mac OS/X. > 
> > > src/serversocket.cpp needs to include <sys/types.h> or there will 
> > be > compilation failures in <netinet/in.h> due to in_addr_t and 
> > other types > not being defined.