On Fri, 26 Oct 2012, Marc Dionne wrote:
On Fri, Oct 26, 2012 at 3:45 AM, Simon Wilkinson
<simonxwilkin...@gmail.com> wrote:
I don't have time to investigate a fix for this, but it's caused by
having two distinct declarations for the same uninitialized data
object, which FreeBSD does not allow in kernel code. If someone else
can propose a fix, poke me through Gerrit and I'll review.
With eca07c108cc61fc9a6da720848e1beb78833f670, neither rx_incomingCallQueue or
rx_idleServerQueue are defined in rx_globals.h. In fact, the only reference to
rx_incomingCallQueue in the tree is in rx.c.
I'm somewhat puzzled by this build failure, unless the FreeBSD builder isn't
cleaning its tree properly between builds?
That would be my guess. Looking at the buildbot log, neither rx.o or
rx_globals.o are rebuilt during the failed build. So we probably have
an rx_globals.o that dates from before that change, while rx.o was
rebuilt after the change.
Hmm, I thought that the builders were supposed to 'git clean -xfd' before
each build?
The kernel makefiles dont't have very detailed dependencies, so they
don't catch that rx_globals.o needs to be rebuilt in this case because
of the cahnge in rx_globals.h.
Hmm, this looks to be using bsd.kmod.mk, we probably need some magic to
set HDRS as well as OBJS.
-Ben
_______________________________________________
OpenAFS-devel mailing list
OpenAFS-devel@openafs.org
https://lists.openafs.org/mailman/listinfo/openafs-devel
