On Mon, Jun 27, 2016 at 05:29:59PM +0100, Wiles, Keith wrote: > On 6/27/16, 7:58 AM, "dev on behalf of Wiles, Keith" <dev-bounces at dpdk.org > on behalf of keith.wiles at intel.com> wrote: > > > > >On 6/27/16, 3:46 AM, "Richardson, Bruce" <bruce.richardson at intel.com> > >wrote: > > > >>On Sun, Jun 26, 2016 at 10:54:12AM -0500, Keith Wiles wrote: > >>> Latest clang compiler 3.8.0 on latest update of Ubuntu > >>> creates a few more warnings on -Warray-bounds and extra > >>> () around 'if' expressions. > >>> > >>> Signed-off-by: Keith Wiles <keith.wiles at intel.com> > >>> --- > >>> app/test-pmd/Makefile | 3 +++ > >>> app/test/Makefile | 3 +++ > >>> drivers/net/bonding/Makefile | 4 ++++ > >>> drivers/net/fm10k/Makefile | 2 ++ > >>> drivers/net/i40e/Makefile | 2 ++ > >>> lib/librte_cmdline/Makefile | 6 ++++++ > >>> lib/librte_eal/linuxapp/eal/Makefile | 8 ++++++++ > >>> 7 files changed, 28 insertions(+) > >>> > >>All the fixes in this patch seem to be just disabling the compiler > >>warnings, which > >>should really be the last resort in cases like this. Can some of the issues > >>be > >>fixed by actually fixing the issues in the code? > > > >I did look at the code to fix the problem, because I could not see one: > > > >/work/home/rkwiles/projects/intel/dpdk/app/test-pmd/cmdline.c:3357:2140: > >error: array index 3 is past the end of the array (which contains 3 > >elements) [-Werror,-Warray-bounds] > > if (!__extension__ ({ size_t __s1_len, __s2_len; (__builtin_constant_p > > (res->proto) && __builtin_constant_p ("ip") && (__s1_len = __builtin_strlen > > (res->proto), __s2_len = __builtin_strlen ("ip"), (!((size_t)(const void > > *)((res->proto) + 1) - (size_t)(const void *)(res->proto) == 1) || __s1_len > > >= 4) && (!((size_t)(const void *)(("ip") + 1) - (size_t)(const void > > *)("ip") == 1) || __s2_len >= 4)) ? __builtin_strcmp (res->proto, "ip") : > > (__builtin_constant_p (res->proto) && ((size_t)(const void *)((res->proto) > > + 1) - (size_t)(const void *)(res->proto) == 1) && (__s1_len = > > __builtin_strlen (res->proto), __s1_len < 4) ? (__builtin_constant_p ("ip") > > && ((size_t)(const void *)(("ip") + 1) - (size_t)(const void *)("ip") == 1) > > ? __builtin_strcmp (res->proto, "ip") : (__extension__ ({ const unsigned > > char *__s2 = (const unsigned char *) (const char *) ("ip"); int __result = > > (((const unsigned char *) (const char *) (res->proto))[0] - __s2[0]); if > > (__s1_len > 0 && __result == 0) { __result = (((const unsigned char *) > > (const char *) (res->proto))[1] - __s2[1]); if (__s1_len > 1 && __result == > > 0) { __result = (((const unsigned char *) (const char *) (res->proto))[2] - > > __s2[2]); if (__s1_len > 2 && __result == 0) __result = (((const unsigned > > char *) (const char *) (res->proto))[3] - __s2[3]); } } __result; }))) : > > (__builtin_constant_p ("ip") && ((size_t)(const void *)(("ip") + 1) - > > (size_t)(const void *)("ip") == 1) && (__s2_len = __builtin_strlen ("ip"), > > __s2_len < 4) ? (__builtin_constant_p (res->proto) && ((size_t)(const void > > *)((res->proto) + 1) - (size_t)(const void *)(res->proto) == 1) ? > > __builtin_strcmp (res->proto, "ip") : (- (__extension__ ({ const unsigned > > char *__s2 = (const unsigned char *) (const char *) (res->proto); int > > __result = (((const unsigned char *) (const char *) ("ip"))[0] - __s2[0]); > > if (__s2_len > 0 && __result == 0) { __result = (((const unsigned char *) > > (const char *) ("ip"))[1] - __s2[1]); if (__s2_len > 1 && __result == 0) { > > __result = (((const unsigned char *) (const char *) ("ip"))[2] - __s2[2]); > > if (__s2_len > 2 && __result == 0) __result = (((const unsigned char *) > > (const char *) ("ip"))[3] - __s2[3]); } } __result; })))) : > > __builtin_strcmp (res->proto, "ip")))); })) { > > > >Here is the line of code for that one: > > if (!strcmp(res->proto, "ip")) { > > > >The ?Wno-parenthese-equality problem gives the output here: > > > >/work/home/rkwiles/projects/intel/dpdk/lib/librte_cmdline/cmdline_cirbuf.c:288:19: > > error: equality comparison with extraneous parentheses > >[-Werror,-Wparentheses-equality] > > if (((cbuf)->len == 0)) { > > > >The line is: > > > > if (CIRBUF_IS_EMPTY(cbuf)) { > > > >This one is in cmdline_cirbuf.h, which can be changed, but I do not think we > >need to remove the parenthese. > > > >I will look at some of other solution, so I rejected the patch. > > I found the problem to the compile errors I am seeing with building with > clang and shared libraries. > > The x86_64-linux-gnu/bits/string2.h header file if getting included from > string.h, but this would be mean __GNUC__ is defined and this is the clang > compiler. After much investigation it turns out ?ccache? is the problem here. > If ccache is enabled with clang builds the __GNUC__ is defined some how, I > never did find the location. > > Just a warning it appears ?ccache? for caching object files is not compatible > with DPDK builds ? in all cases. > Actually, I believe it's a more general ccache and clang problem, not DPDK specific.
See e.g. http://petereisentraut.blogspot.com/2011/09/ccache-and-clang-part-2.html where the recommendation is to set "export CCACHE_CPP2=yes" in your environment. This cleared quite a number of issues for me (and others) when compiling with clang. [Credit too to Ferruh who first pointed this issue out to me] /Bruce