On 10/31/2016 06:01 PM, Sven C. Dack wrote:
On 31/10/16 17:16, Nikolaus Dunn wrote:
...
Both libA and libB have the symbol:
_ZNSt15basic_stringbufIcSt11char_traitsIcE18tracking_allocatorIcEE8overflowEi
How do I get libA to use its own version of that symbol? I thought
that was what -Bsymbolic-functions was supposed to do.
Thanks,
Nik
I don't think you can have the same symbol twice in a symbol table.
-Bsymbolic only protects against symbol overwriting by the executable,
but you still have the issue of having multiple definitions for the
same symbol. You will have to use "-z muldefs" if you don't already
use it, which however will only allow for multiple definitions, but
the linker will then pick the first definition it finds and discards
all the others. So I'm not sure if this is possible at all.
You could try to undefine the symbol with "-u name" and thereby avoid
it from being entered into the global symbol table. Or undefine the
symbol for each libA and liB just before you link against them.
Thank you for your response. I think the issue I'm having is because
although I've statically linked against libstdc++, there are vtables
with the same name in more than one library. I suspect the application
is picking the first vtable that matches the class name and using that.
If so, is this behavior customizable?
If library A defines a class Foo and library B defines a class Foo and
the two are used in the same application, would it be possible to force
any code in library A that instantiates instances of Foo to use library
A's definition of the vtable and any code in library B that does the
same to use its definition of the vtable for Foo?
Is this possible?
I can't find much information on how the vtables themselves are loaded
in an application and how the value for vptr is established.
Thanks,
Nik
