On Monday 19 November 2007 03:55:20 Steven Edwards wrote:
The headers are not an issue because we use the Wine headers and import libraries for the dlls. The only place there is a problem is with wine/programs and the regression tests. They use the native import libraries. I don't really understand Alexandre's reasoning on this as myself and Dmitry have both sent patches in to correct this issue and his answer has always been that we should fix mingw w32api. Wine always tends to break when cross-compiling because of this. If we always defaulted to using the Wine headers and Import libs there never would be a problem in that regards.
I recall Alexandre saying that he wants the exposure to alternative headers and import libraries because it brings up bugs in Wine. He also suggested that the wininet/winsock issue could be solved by making wininet build against either Windows sockets or Unix sockets.
Here's a historical thread on this issue:
http://www.kernel-traffic.org/wine/wn20040416_219.html#5
-Hans