Ubuntu
linux package

  1. Bionic (18.04)
  2. Bug #1823429
  3. Comment #2

Comment 2 for bug 1823429

Revision history for this message
Dan Streetman (ddstreet) wrote : Re: linux-oracle ftbfs in disco

the same 4.15 kernel is getting built on b/c/d/e, and scripts/selinux/genheaders/genheaders.c includes the local header files, specifically <sys/socket.h>:
#include <sys/socket.h>
which define PF_MAX (via AF_MAX) as 45 for cosmic and later. however genheaders.c also:
#include "classmap.h"
which is security/selinux/include/classmap.h and that does:
#if PF_MAX > 44
#error New address family defined, please update secclass_map.
#endif

$ git grep AF_MAX ubuntu/bionic/master-next socket.h
ubuntu/bionic/master-next:socket.h:#define AF_MAX 44 /* For now.. */
ubuntu/bionic/master-next:socket.h:#define PF_MAX AF_MAX
$ git grep AF_MAX ubuntu/cosmic/master-next socket.h
ubuntu/cosmic/master-next:socket.h:#define AF_MAX 45 /* For now.. */
ubuntu/cosmic/master-next:socket.h:#define PF_MAX AF_MAX
$ git grep AF_MAX ubuntu/disco/master-next socket.h
ubuntu/disco/master-next:socket.h:#define AF_MAX 45 /* For now.. */
ubuntu/disco/master-next:socket.h:#define PF_MAX AF_MAX
$ git grep AF_MAX ubuntu/eoan/master-next socket.h
ubuntu/eoan/master-next:socket.h:#define AF_MAX 45 /* For now.. */
ubuntu/eoan/master-next:socket.h:#define PF_MAX AF_MAX

This is probably a bug that should be fixed in genheaders.c upstream, although I don't quite understand why the 4.15 linux-oracle kernel is getting built on cosmic, disco, or eoan.