On 26 February 2015 at 21:26, Chad Versace <[email protected]> wrote: > On 02/23/2015 12:32 PM, Emil Velikov wrote: >> Hello list, >> >> Second to last part of removing all the window specific bits from >> link-time to run-time (hail libdl). >> >> Note that the last patch looks a bit messy, mostly due to the wayland >> design approach to have a handful of functions defined as static inlines >> in the public header. >> >> This has been compiled tested only, as I don't have a wayland setup atm. >> Will see if anyone volunteers over at #wayland (touch wood). > > I run Waffle's Wayland tests inside an X session. You can run Weston > inside an X window. > > $ sudo yum install weston > $ nohup weston & > $ wflinfo -p wayland -a gles2 > >> >> As usual the series can be found at my github repo in branch >> wip/no-libwayland. >> >> Comments, suggestions even moderate bike-shed is welcome. > > wl_display_connect() crashes with SIGSEGV. I suspect the problem > is a symbol collision: a wl_display_connect symbol exists in > libwaffle and in the dlopened libwayland-client. But I'm not > confident in that guess. > > This patch series is crazy stuff :) Before it's complete > and working, we'll probably learn a few things about linker magic. > > Maybe I should be building libwaffle with -Bsymbolic? Hmm.... > I fear that -Bsymbolic does not cut it :'(
I've went ahead with another solution inspired by SDL. Admittedly it's not pretty, but it does not violate the spec (like v1), plus it even works - double-checked via your instructions above. The whole things in a nutshell can be seen below: $ cat wrapper.h foo * (*wrapper_wl_foo)(...); #define wl_foo (*wrapper_wl_foo) #include <wayland-foo.h> We're nearly there :-) -Emil _______________________________________________ waffle mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/waffle
