dbus remaining windows port issues was Re: moving file descriptor passing into a header

Fridrich Strba fridrich.strba at bluewin.ch
Thu Apr 22 03:18:13 PDT 2010


My dearest :)

On Thu, 2010-04-22 at 10:44 +0300, Tor Lillqvist wrote: 
> Is the current situation such that using a configure && make way to
> build is totally broken if building for Windows? (Or maybe just if
> cross-compiling for Windows?) That's what I used last time I built
> dbus master on Windows, but yeah, it certainly was not
> straightforward, and I don't know/remember if any required changes to
> the configury/makefilery have been upstreamed.

No, the configure build is actually working just fine right now with the
recent master after some fixes went in. I just needed to build the cmake
build to be sure that my de-kde-ification patch did not break anything
there and also run the unit tests there. And since I don't use windows
for builds, I cross-compiled it and then run the tests in a vmware where
the unix file-system was a shared folder. That is where I got these two
issues and considered worthy to report in case someone knows how to
handle the cross-compiling situation the standard way with cmake. The
macro I attached is a macro that I use for some libraries/applications
in my repositories where there is no other choice but to use it and it
works generally well, but in this case.

> (Also, as far as I recall, some macro assignments passed on the
> command line differ in the cmake and a "normal" configure&&make build,
> so there might be an even higher risk of non-interoperability with the
> cmake/"kde" style build on Windows then...)

Yeah, apart the directory layout that is a bit different, I was just
wondering whether it would not be worthy to check and unify other things
that can be done. Like which socket is used in cmake and in autofoo
build ...

Cheers

Fridrich




More information about the dbus mailing list