Re: [PATCH v2] mk: optimize directory dependencies

[Olivier Matz <olivier.matz@6wind.com>]

On Thu, 30 Mar 2017 10:27:49 +0100, Ferruh Yigit <ferruh.yigit@intel.com> wrote:
> On 3/30/2017 9:51 AM, Olivier Matz wrote:
> > Hi Ferruh,
> > 
> > On Tue, 28 Mar 2017 11:34:35 +0100, Ferruh Yigit <ferruh.yigit@intel.com> wrote:  
> >> On 3/24/2017 1:21 PM, Olivier Matz wrote:  
> >>> Before this patch, the management of dependencies between directories
> >>> had several issues:
> >>>
> >>> - the generation of .depdirs, done at configuration is slow: it can take
> >>>   more than one minute on some slow targets (usually ~10s on a standard
> >>>   PC without -j).
> >>>
> >>> - for instance, it is possible to express a dependency like:
> >>>   - app/foo depends on lib/librte_foo
> >>>   - and lib/librte_foo depends on app/bar
> >>>   But this won't work because the directories are traversed with a
> >>>   depth-first algorithm, so we have to choose between doing 'app' before
> >>>   or after 'lib'.
> >>>
> >>> - the script depdirs-rule.sh is too complex.
> >>>
> >>> - we cannot use "make -d" for debug, because the output of make is used for
> >>>   the generation of .depdirs.
> >>>
> >>> This patch moves the DEPDIRS-* variables in the upper Makefile, making
> >>> the dependencies much easier to calculate. A DEPDIRS variable is still
> >>> used to process library dependencies in LDLIBS.
> >>>
> >>> After this commit, "make config" is almost immediate.
> >>>
> >>> Signed-off-by: Olivier Matz <olivier.matz@6wind.com>
> >>> Tested-by: Robin Jarry <robin.jarry@6wind.com>
> >>> Tested-by: Jerin Jacob <jerin.jacob@caviumnetworks.com>    
> >>
> >> <...>
> >>  
> >>>  DIRS-$(CONFIG_RTE_LIBRTE_PORT) += librte_port
> >>> +DEPDIRS-librte_port := librte_eal librte_mbuf librte_mempool librte_ether
> >>> +DEPDIRS-librte_port += librte_ip_frag librte_sched librte_kni    
> >>                                                          ^
> >> ---------------------------------------------------------+
> >>
> >> <...>
> >>  
> >>> diff --git a/lib/librte_port/Makefile b/lib/librte_port/Makefile
> >>> index 44fa73520..76629a13a 100644
> >>> --- a/lib/librte_port/Makefile
> >>> +++ b/lib/librte_port/Makefile    
> >> <...>  
> >>> -ifeq ($(CONFIG_RTE_LIBRTE_KNI),y)
> >>> -DEPDIRS-$(CONFIG_RTE_LIBRTE_PORT) += lib/librte_kni
> >>> -endif    
> >>
> >> This check needs to be reflected to new method, otherwise causing build
> >> errors for i686 target.
> >>  
> > 
> > Did you notice an error? I see no error it with:
> > 
> >     make config T=x86_64-native-linuxapp-gcc
> >     sed -i 's,CONFIG_RTE_LIBRTE_KNI=y,# CONFIG_RTE_LIBRTE_KNI is not set,' build/.config
> >     make -j8
> > 
> > 
> > I think it is not a problem, the DEPDIRS-* variable only gives an
> > ordering directive, which should be ok even if kni is not compiled.  
> 
> Can you please test shared library build, I am getting following error:
> 
> == Build lib/librte_port
>   LD librte_port.so.3.1
> /usr/bin/ld: cannot find -lrte_kni
> collect2: error: ld returned 1 exit status
> .../mk/rte.lib.mk:127: recipe for target 'librte_port.so.3.1' failed
> make[3]: *** [librte_port.so.3.1] Error 1
> .../mk/rte.subdir.mk:63: recipe for target 'librte_port' failed
> 

You're right, it happens because the libray list is built from DEPDIRS.

Thanks for spotting it, I'll check if there are other occurences and send
a fix. The simple way is probably to restore the ifeq(), but I'm wondering
if having: DEPDIRS-$(cond)-$(path) wouldn't be shorter in Makefiles.

Olivier