Re: [Gnash-dev] how to handle hash_map

[Erik de Castro Lopo]

On Mon, 27 Feb 2006 09:24:36 -0700
Rob Savoye <address@hidden> wrote:

>    I believe all the VC++ compilation support has likely bit-rotted. It 
> used to be built in using the Cygwin make, with a command line VC++.

I have a Free Software project in C (libsndfile in case anyone has heard
of it) which is quite widely used on windows. For many years I spent
considerable amounts of time attempting to support VC++ (verisons 5 through
8). Build issues around VC++ seemed to be about 70% of my libsndfile
related email.

I recently gave up supporting VC++ and started distributing a windows
ZIP file containing MinGW compiled binaries and all the source code.
The MinGW compiled binaries are generated using the same configure
scripts and makefile used to build the project on unix.

Since switching to this binary plus source distribution for libsndfile,
I have not received one email complaining about lack of VC++ support.
Anyone that does complain will be directed to (hmmm, i've just found a 
typo :-)):

    http://www.mega-nerd.com/libsndfile/win32.html

>    For win32, I believe Firefox does use VC++, so a plugin should too. 
> I'd prefer sticking to GCC where possible, but somehow I doubt Firefox 
> uses Cygwin or MingW.

I presume the Firefox plugin is just a DLL. If that DLL only requires
C linkage (ie no C++ name mangling) then MinGW should be fine. Linking
programs compiled with VC++ to C linkage DLLs compiled with MinGW works.

The problem with C++ linkage is that VC++ and GCC use different name 
mangling schemes.

>    It would be nice to find a "clean" win32 developer that could take on 
> maintaining Gnash on that platform...

That person must also be an active developer because its surprising how
quickly the autoconf/automake based configuration can leave a stale
windows build system behind.

Erik
-- 
Erik de Castro Lopo
Senior Design Engineer
Start Corporation
http://www.startcorp.com