Ben Hutchings wrote...
> radvd's autoconf test for <linux/if.h> has probably failed at least
> since Linux 2.6.32 when I made sure the kernel headers would never
> define struct sockaddr for userland:
> <https://git.kernel.org/linus/9c501935a3cdcf6b1d35aaee3aa11c7a7051a305>
>
> But the conflict between <linux/if_arp.h> and <net/if_arp.h> is far
> older than that, so if the test ever passed it should have resulted in
> this build failure. I think that's a clear bug in radvd. It should
> use either one or the other, and I think the sensible thing is to use
> <net/if_arp.h> as it has been doing up until now.
Certainly I'm not going to advocate radvd's build system - my only
concern is your change will break more packages out there, mere ten days
prior to the scheduled release.
So to me it seems wiser to revert the changes to linux-libc-dev for
stretch. Else, someone(tm) would have to test-rebuild all the packages
that build-depend on linux-libc-dev to check for damage, and provide
help how to deal with something that technically is a library
transition.
Christoph
Attachment:
signature.asc
Description: Digital signature