[cfe-dev] Symbol names for constructors

[Tamas Berghammer]

We try to find a function based on the mangled symbol name and it fails as
they do not match (looking for _ZN5myIntC1Ei while the executable
contains _ZN5myIntC2Ei). We can do a search based on de-mangled name but if
the executable will contain both _ZN5myIntC1Ei and _ZN5myIntC2Ei then we
will find both of them as they de-mangle to the same name (myInt::myInt(int))
because the only difference between them is that one of them should be
called when constructing an object of type myInt (_ZN5myIntC1Ei) and the
other one should be called when constructing an object with base class of
myInt (_ZN5myIntC2Ei).

On Wed, Apr 8, 2015 at 12:46 PM, Azat Khuzhin <a3at.mail at gmail.com> wrote:

> On Wed, Apr 08, 2015 at 12:21:23PM +0100, Tamas Berghammer wrote:
> > I am compiling the following code with clang TOT (clang version 3.7.0
> > (trunk 234376) (llvm/trunk 234392) Target: x86_64-unknown-linux-gnu
> Thread
> > model: posix) with the following command:
> >
> > === source ===
> > class myInt {
> > public: myInt(int _x) : theValue(_x) {}
> > private: int theValue;
> > };
> >
> > int main() {
> >     myInt x{3};
> >     return 0;
> > }
> >
> > === compile command ===
> > ./bin/clang++ --std=c++11 ~/tmp/C12.cpp
> >
> > After compilation I see a symbol created for the constructor with name
> > "_ZN5myIntC2Ei" what is the base object constructor but I don't have a
> > symbol for "_ZN5myIntC1Ei" what would be the complete object constructor.
>
> You need to demangle symbol, since C++ do name mangling to support
> functions with identical names and different arguments for example.
>
> You could use c++filt for this:
> $ c++filt _ZN5myIntC2Ei
> myInt::myInt(int)
>
> > This issue cause us some problem in LLDB during expression evaluation so
> I
> > am interested if it is a bug (and clang should generate both symbol as
> gcc
> > do) or it is an intended behavior as an optimization. In either case I
> > would appreciate if someone can give me an idea about what is the case
> when
> > only one of the symbol is emitted so I can handle that case in LLDB
> > properly.
>
> What problem in lldb?
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20150408/57892450/attachment.html>