[Zlib-devel] zlib 1.2.3.9 is out (fwd)

[Vincent Torri]

Hey

On Mon, 22 Feb 2010, Mark Adler wrote:

> On Feb 21, 2010, at 9:56 PM, Vincent Torri wrote:

> A problem that has been seen in previous DLLs is that not all of the 
> external symbols that are supposed to be available are.  What file is 
> used to determine that, so that I can verify that they are all there?

With MSYS and MinGW, you can list the symbols from the static lib with nm, 
like on linux:

nm -C libz.a

and check if they are defined or not (no 'U' before). Is it what you want 
?

>> 3) Pedro mentioned that the name of the import library should be reverted (because of change in the way people linked against zlib before the change). I don't know if you agree with him or not.
>
> I like libzdll.a better, but I'm not hard over either way since I don't 
> live in that environment.  Is there a consensus as to what it should be 
> named?

  * libtool uses by default the following convention:

foo**.dll      the DLL
libfoo.dll.a   the import library
libfoo.a       the static library

and to link against foo**.dll, pass to the linker -lfoo

note: visual studio names the import library foo.lib

  * it seems that for zlib, it was:

zlib1.dll  the DLL
libzdll.a  the import library
libz.a     the static library

and to link against zlib1.dll, pass to the linker -lzdll

I don't know if there is a concensus, it's just a matter of use. As Pedro 
stated, after i suggested you to change the name, all the previous 
configuration of programs that use zlib might not link against the dll. So 
maybe it is more safe to revert the change of the name.

> I'm not too fond of the #cmakedefines in zconf.h.in.  Is there a way to 
> get CMake to put those in, as opposed to depending on ./configure to 
> take them out?

that would be better, indeed.

Vincent Torri