[PATCH] BZ #17702: Fix recursive dlopen.

[Carlos O'Donell]

On 12/12/2014 08:33 PM, Carlos O'Donell wrote:
> On 12/12/2014 05:27 PM, Andreas Schwab wrote:
>> "Carlos O'Donell" <carlos@redhat.com> writes:
>>
>>> On 12/12/2014 06:10 AM, Andreas Schwab wrote:
>>>> "Carlos O'Donell" <carlos@redhat.com> writes:
>>>>
>>>>> +LDLIBS-tst-rec-dlopen = -ldl
>>>>> +$(objpfx)tst-rec-dlopen: $(libdl)
>>>>
>>>> Why do you need to set LDLIBS?
>>>
>>> How else do I get it to to link libdl.so against the test case?
>>
>> It's listed as a dependency.
> 
> I see. It does indeed work with just the dependency.
> I guess I don't know what LDLIBS is for? Is it only
> for building DSOs?
> 
> Here is a v2 with several fixes, including checking to make sure
> that the return value of the called function is indeed that of
> moddummy1 and provides some assurance that nothing failed.
> 
> Cheers,
> Carlos.
> 
> v2.
> - Remove LDLIBS
> - tst-rec-dlopen should depend on moddummy1 and moddummy2 DSOs.
> - Use return of DSO to determine if DSO func was called and that
>   dummy1 was the outermost called function.

I have committed v2 to allow dlopen to be called recursively.

We're using this in RHEL to allow 3rd party malloc interposes
to use dlopen.

Cheers,
Carlos.