We need to add configure logic so that we can build the Windows Media Foundation backend I'm making in bug 799315. We'll build it by default on Windows, and pref it on (probably preffing it off on Aurora if it's not ready to ship). We'll want an option to disable building the backend too.

Build Windows Media Foundation backend, and pref the backend on, on Windows.

Comment on attachment 685770 [details] [diff] [review] Patch: build system changes >+#ifdef XP_WIN >+pref("media.windows-media-foundation.enabled", true); >+#endif Why is XP_WIN used here instead of MOZ_WMF?

Good point. I will make it MOZ_WMF.

Comment on attachment 685770 [details] [diff] [review] Patch: build system changes Review of attachment 685770 [details] [diff] [review]: ----------------------------------------------------------------- ::: configure.in @@ +5349,5 @@ > + dnl Enable Windows Media Foundation support by default. > + dnl Note our minimum SDK version is Windows 7 SDK, so we are (currently) > + dnl guaranteed to have a recent-enough SDK to build WMF. > + MOZ_WMF=1 > +fi Hmm, should we turn this off by default if doing mingw builds? I guess we could just wait for Jacek to file the bug saying we broke him :-P @@ +5357,5 @@ > + MOZ_WMF=, > + MOZ_WMF=1) > + > +if test -n "$MOZ_WMF"; then > + AC_DEFINE(MOZ_WMF) Some bizarre tab spacing here? @@ +5360,5 @@ > +if test -n "$MOZ_WMF"; then > + AC_DEFINE(MOZ_WMF) > + MOZ_SYDNEYAUDIO=1 > + MOZ_CUBEB=1 > + MOZ_MEDIA=1 Extra whitespace at EOL.

Jacek: Do you have any input as to how my patch here can not break on your mingw builds?

As far as build system changes are concerned, the patch looks good to me. If there is a problem with compiling the code, I'd much rather fix the code than disable it by default (and meanwhile --disable-wmf should work as a workaround). BTW, quickly looking at the code (the version I found in bug 799315 try commit), it will require adding some missing bits to upstream mingw-w64 and at least trivial case sensitivity (platform includes should be lower case) and #pragma comment(lib, ...) (replacing them with linker arguments) changes, but I may do that after the code lands.

OK, thanks. The most recent version of my patches are in this try push: https://tbpl.mozilla.org/?tree=Try&rev=f5b02478ecb0 I switched from using static linking to dynamic linking, so hopefully you won't need to worry about linker arguments etc.

Landed, but preff'd off: https://hg.mozilla.org/integration/mozilla-inbound/rev/126256fe6355

https://hg.mozilla.org/mozilla-central/rev/126256fe6355 https://hg.mozilla.org/mozilla-central/rev/7f71d8e13583