whatis DYNAMIC_NAME_CLASH_OK

whatis DYNAMIC_NAME_CLASH_OK

[jarausch]

Hi,

I've just debugged a segment fault, when trying to load
Mike Fletcher's zperl module.

I traced it down to 
#0  module_func (m=0x100da3a8, name=0x100ad0f8 "setup_", 
    name_s=0x100ad100 "setup_%s") at module.c:468
#1  0x1005746c in setup_module (m=0x100da3a8) at module.c:482



Now in  'module_func) (module.c:452)

s= m->nam  returning  "zperl"

since configure defined DYNAMIC_NAME_CLASH_OK on my IRIX 6.5
it directly calls
fn = (Module_func) dlsym(m->handle, name);
where name = 'setup_'
Obviously that fails (since the correct symbol is  setup_zperl)
and the success of dlsym is not checked
we run into nirwana.

1.) What's the heck with DYNAMIC_NAME_CLASH_OK ?
2.) At least IRIX 6.5 obviously needs  #undef DYNAMIC_NAME_CLASH_OK
3.) checking the success of  dlsym and an error abort
    would be nicer than a segment fault, which was hard to be traced
    down

Thanks for your comments,
Helmut.

-- 
Helmut Jarausch
Lehrstuhl fuer Numerische Mathematik
Institute of Technology, RWTH Aachen
D 52056 Aachen, Germany

Re: whatis DYNAMIC_NAME_CLASH_OK

[Zefram]

jarausch@igpm.rwth-aachen.de wrote:
>1.) What's the heck with DYNAMIC_NAME_CLASH_OK ?

That's from a configure test, which checks whether you can load multiple
modules that define symbols with the same name.  The reason for it is
that, on systems that permit it, all setup functions are named "setup_"
(and the other module functions similarly), so that a module works
independently of its filename.

-zefram

Re: whatis DYNAMIC_NAME_CLASH_OK

[jarausch]

On 26 Aug, Zefram wrote:
> jarausch@igpm.rwth-aachen.de wrote:
>>1.) What's the heck with DYNAMIC_NAME_CLASH_OK ?
> 
> That's from a configure test, which checks whether you can load multiple
> modules that define symbols with the same name.  The reason for it is
> that, on systems that permit it, all setup functions are named "setup_"
> (and the other module functions similarly), so that a module works
> independently of its filename.
> 
> -zefram

Thanks,
now I have to find out why dlsym returns an invalid address when
called with 'setup_' only (or should it be 'setup' ?)

Helmut.


-- 
Helmut Jarausch
Lehrstuhl fuer Numerische Mathematik
Institute of Technology, RWTH Aachen
D 52056 Aachen, Germany