30/01/2018 10:09, gowrishankar muthukrishnan:
> On Tuesday 30 January 2018 02:26 PM, Jerin Jacob wrote:
> > -----Original Message-----
> >> Date: Tue, 30 Jan 2018 14:18:46 +0530
> >> From: Gowrishankar <gowrishanka...@linux.vnet.ibm.com>
> >> To: dev@dpdk.org
> >> CC: Olivier Matz <olivier.m...@6wind.com>, tho...@monjalon.net,
> >> Gowrishankar Muthukrishnan <gowrishanka...@linux.vnet.ibm.com>
> >> Subject: [dpdk-dev] [PATCH] ring: fix compilation error with a broken else
> >> clause
> >> X-Mailer: git-send-email 1.9.1
> >>
> >> From: Gowrishankar Muthukrishnan <gowrishanka...@linux.vnet.ibm.com>
> >>
> >> Calling rte_smp_wmb macro expands into a compound block, which
> >> would break compiling a else clause following it, if that calling
> >> place has been terminated already with ";", as in below code.
> >> This patch adds { } around this macro to allow compiling else too.
> >>
> >> Fixes: git show c9fb3c6289 ("ring: move code in a new header file")
> >>
> >> Signed-off-by: Gowrishankar Muthukrishnan
> >> <gowrishanka...@linux.vnet.ibm.com>
> >> ---
> >>
> >> Error in compiling source:
> >>
> >> In file included from /tmp/dpdk/lib/librte_ring/rte_ring.h:372:0,
> >> from /tmp/dpdk/lib/librte_ring/rte_ring.c:90:
> >> /tmp/dpdk/lib/librte_ring/rte_ring_generic.h: In function ‘update_tail’:
> >> /tmp/dpdk/lib/librte_ring/rte_ring_generic.h:75:2: error: ‘else’ without a
> >> previous ‘if’
> >> else
> >> ^~~~
> >> /tmp/dpdk/mk/internal/rte.compile-pre.mk:114: recipe for target
> >> 'rte_ring.o' failed
> >>
> >> lib/librte_ring/rte_ring_generic.h | 3 ++-
> >> 1 file changed, 2 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/lib/librte_ring/rte_ring_generic.h
> >> b/lib/librte_ring/rte_ring_generic.h
> >> index 8c3e65b..a668489 100644
> >> --- a/lib/librte_ring/rte_ring_generic.h
> >> +++ b/lib/librte_ring/rte_ring_generic.h
> >> @@ -70,8 +70,9 @@
> >> update_tail(struct rte_ring_headtail *ht, uint32_t old_val, uint32_t
> >> new_val,
> >> uint32_t single, uint32_t enqueue)
> >> {
> >> - if (enqueue)
> >> + if (enqueue) {
> > Just wondering, Is it because of rte_smp_wmb() implementation of ppc64 is
> > not just correct?
> >
> > We had a similar fix for arm64.
> > http://dpdk.org/browse/dpdk/commit/?id=59a3cae5305816b3739b4197d277f9455af53b9f
> Thanks Jerin. Yes. Similar fix also works. I did not want to touch
> include/ at this time, as it is called at various places in source.
> As same problem is seen in arm64, I think it is ok to go with your
> change. If so, I can send version 2 patch with similar change as you showed.
Please send a patch for the fix in atomics.
