dev-cpp-users

[Dev-C++] problem using dll from visual studio

[Axel B. <new...@th...>]

Hi Folks,

I'm new to the list because I couldn't find a solution to this little 
problem.

I am trying to use the sleepycat berkeley db 4.2, compiled under visual 
studio 6, with dev-c++. but this won't work - when linking I always get 
the following error:

	[Linker error] undefined reference to `Db::Db(DbEnv*, unsigned)'

this bugs me. I tried to include the .lib-file by -l switch, directly, I 
set the include/library paths, I tried to include the .a file, etc. 
nothing works.

AND I checked with nm that the lib indeed HAS the functions included - 
the cxx_*.obj-files are present in the lib. so - I'm clueless :-/

can anyone please help me, its kind of important ... :-)


thanks in advance & greetings,

axel.

[Dev-C++] problem using dll: is c++ error??!?

[Axel B. <new...@th...>]

Ok, this is the last email I'll write. I have figures something out: I 
simply tried to call a C (!) API function from my program, and - 
wo-hooo! - the linker did as told. ONLY the constructor of the OBJECT 
(C++-API of berkeley) seems not to work.

I didnt try whether the complete C API is present and working, I just 
tried to call a function and to compile, but this seems very strange.

I'm sorry if I'm writing too much, but this did cost me the better half 
of this afternoon, and I don't understand it.

But - and I checked that again - all the C++ functions *seem* to be 
present in the library ... .


Again: any ideas? :-))


last time thanks and greetings,

axel.


p.s.: if I should solve this (not likely) I will post the solution, too ;-)

Re: [Dev-C++] problem using dll: is c++ error??!?

[<gen...@ya...>]

axel, there's comething called 'name mangling' that a c++ compiler does to c++ functions. you can't use a c++ function call compiled in VC++ from dev-c++, because the compilers use different name-mangling.

 

regards

		
---------------------------------
  Yahoo! Messenger - Communicate instantly..."Ping" your friends today! Download Messenger Now

Re: [Dev-C++] problem using dll: is c++ error??!?

[Axel B. <new...@th...>]

Oladipo Odumosu wrote:
> axel, there's comething called 'name mangling' that a c++ compiler does 
> to c++ functions. you can't use a c++ function call compiled in VC++ 
> from dev-c++, because the compilers use different name-mangling.

wow. thank you - thats a hint at least .... . google is failing on this
matter ;-) I figured something like this by myself in the meantime, but
I had no clue - it was basically just a thought, and this gives some
substance to it.

BUT - du you have a solution? can you switch the name mangling from vc++
to the dev-c++ model? is it *somehow* possible that this will ever work?
  can you convert the dll/lib/whatever, can you tell dev-c++ how to do it?


thanks again,

axel.

Re: [Dev-C++] problem using dll: is c++ error??!?

[<or...@vp...>]

Axel Bock wrote:
> BUT - du you have a solution? can you switch the name mangling from vc+=
+
> to the dev-c++ model? is it *somehow* possible that this will ever work=
?
>  can you convert the dll/lib/whatever, can you tell dev-c++ how to do i=
t?

No, sadly that will never be possible in the foreseeable future.

To complicate things even more, GCC itself changes it's own name=20
mangling-scheme from time-to-time. That means for example, C++ libraries=20
(not just DLLs, but static ones also) that are compiled with GCC 3.2.3=20
cannot be used with GCC 3.3.3, because of the name mangling changed=20
between these versions too.
I just added this as a side note, because it caused quite a lot of=20
headache for me until I figured this out... :)

--
Greetings,
   Bal=E1zs

Re: [Dev-C++] problem using dll: is c++ error??!?

[Axel B. <th...@th...>]

OROSZI Balázs wrote:

> No, sadly that will never be possible in the foreseeable future.
> 
> To complicate things even more, GCC itself changes it's own name 
> mangling-scheme from time-to-time. That means for example, C++ libraries 
> (not just DLLs, but static ones also) that are compiled with GCC 3.2.3 
> cannot be used with GCC 3.3.3, because of the name mangling changed 
> between these versions too.
> I just added this as a side note, because it caused quite a lot of 
> headache for me until I figured this out... :)

oh yeah, thanks, this fact is just sooooooo well documented. if I had 
known this I hadnt even tried ... . does anyone actually mention this in 
its documentation or is this just a fact "one knows"?? :-)

one wonders whether this should be standardized so that ALL libs from 
ALL compilers could be used together. well.

but anyhow - i managed to re-compile berkeley db with dev-c++, using the 
"import msvc project" function (GREEAAAAATTT!!!! - I did not realize I 
had such a thing until it bit me ;-) and modifying some #defines. now it 
should work just fine :-))


thanks and greetings,

axel.

Re: [Dev-C++] problem using dll: is c++ error??!?

[Per W. <pw...@ia...>]

One reason why gcc has changed is name-mangling is that gcc have also
changed it's internal representation of objects and virtual method tables.

It is - of course - only possible to link c++ code from two different
compilers if calling conventions and memory layout of objects etc are
identical betwen the two compilers. Any little difference would just
result in a failing program. Sometimes just generating the wrong answer
but most probably exploding because of a pointer error or similar.

So the real problem isn't name mangling - which would be possible to solve
by C stub functions or some playing around with compiler switches - but
that the compilers must use a common ABI.

See
http://gcc.gnu.org/gcc-3.2/c++-abi.html
for some more information.

/Per W

On Thu, 22 Apr 2004, OROSZI Bal=E1zs wrote:

> Axel Bock wrote:
> > BUT - du you have a solution? can you switch the name mangling from vc+=
+
> > to the dev-c++ model? is it *somehow* possible that this will ever work=
?
> >  can you convert the dll/lib/whatever, can you tell dev-c++ how to do i=
t?
>
> No, sadly that will never be possible in the foreseeable future.
>
> To complicate things even more, GCC itself changes it's own name
> mangling-scheme from time-to-time. That means for example, C++ libraries
> (not just DLLs, but static ones also) that are compiled with GCC 3.2.3
> cannot be used with GCC 3.3.3, because of the name mangling changed
> between these versions too.
> I just added this as a side note, because it caused quite a lot of
> headache for me until I figured this out... :)
>
> --
> Greetings,
>    Bal=E1zs
>
>
> ----