On 10/22/05, Matthew Dillon <[EMAIL PROTECTED]> wrote: > > :Following patch move DEVICE_POLLING from opt_global.h to opt_polling.h: > :http://leaf.dragonflybsd.org/~sephe/opt_polling.diff > : > :Please review it. > : > :Best Regards, > :sephe > : > :-- > :Live Free or Die > > net/if_var.h has all sorts of conditionals on DEVICE_POLLING. Either
net/if_var.h was left out intentionally, since 1) Size and layout of "struct ifnet" are not changed, no matter whether DEVICE_POLLING is defined or not. So any *.c files that does not want DEVICE_POLLING can safely ignore opt_polling.h 2) Any *.c file that wants DEVICE_POLLING supports will include opt_polling.h at its very beginning(at least before net/if.h or net/if_var.h). This will bring in ifnet.if_poll, poll_cmd and various polling related functions' declaration. Again, as pointed out in 1), size and layout of "struct ifnet" are not changed. 3) IMHO, including opt_*.h in system header files will cause trouble > net/if_var.h needs to include opt_polling.h (conditionally on _KERNEL > or KLD_MODULE), or the option has to be left in opt_global.h. > > I think there were some recent commits that allow modules to use > the primary opt_*.h files for a compile so its possible that this > won't break the module build, but it would have to be tested. If module is not built during kernel building, the primary opt_*.h will not be used. This is the main reason that I want to put DEVICE_POLLING in a seperate opt_x.h Best Regards, sephe -- Live Free or Die
