On Thu, Oct 7, 2010 at 9:14 AM, Simon Wilkinson <[email protected]> wrote: > > On 7 Oct 2010, at 08:40, [email protected] wrote: > >> /home/afs/openafs/src/afs/LINUX/osi_compat.h:158: error: implicit >> declaration of function ?key_alloc? > > Your kernel doesn't have keyring support, and it looks like we may have > broken builds in that case. > > You could try adding: "#ifdef LINUX_KEYRING_SUPPORT" at line 148 of that > file, and a "#endif" at line 200, and see if that helps you build further. > > Cheers, > > Simon.
We'll probably need the above for the case where keyring support is not available, but the current setup for the header files could be a problem even if keyrings are enabled. I was intrigued as to how we got key.h included for all files that include osi_compat.h (quite a few), when currrently osi_groups.c is the only one that specifically includes key.h. Looking at a typical pre-processed file on RH5 shows that we get key.h indirectly from security.h and sysincludes.h. The security.h include is conditional on a config test. This seems too fragile and potentially broken depending on the kernel version. Since we need key.h in the keyring case to parse osi_compat.h for all LINUX specific files, it would be better to have key.h included in sysincludes.h (where it used to be), or maybe directly in osi_compat.h. Thoughts? Marc _______________________________________________ OpenAFS-devel mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-devel
