diff mbox

[dpdk-dev,RFC] mk: filter duplicate configuration entries

Message ID 1465830632-25156-1-git-send-email-christian.ehrhardt@canonical.com (mailing list archive)
State Superseded, archived
Delegated to: Thomas Monjalon
Headers

Commit Message

Christian Ehrhardt June 13, 2016, 3:10 p.m. UTC 
  Due to the hierarchy and the demand to keep the base config shoing all
options some options end up multiple times in the .config file.

A suggested solution was to filter for duplicates at the end of the
actual config step which is implemented here.

Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
---
 mk/rte.sdkconfig.mk | 5 +++++
 1 file changed, 5 insertions(+)

Comments

Ferruh Yigit June 28, 2016, 4:11 p.m. UTC | #1 
On 6/13/2016 4:10 PM, Christian Ehrhardt wrote:
> Due to the hierarchy and the demand to keep the base config shoing all
> options some options end up multiple times in the .config file.
> 
> A suggested solution was to filter for duplicates at the end of the
> actual config step which is implemented here.
> 
> Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
> ---
>  mk/rte.sdkconfig.mk | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/mk/rte.sdkconfig.mk b/mk/rte.sdkconfig.mk
> index a3acfe6..734aa06 100644
> --- a/mk/rte.sdkconfig.mk
> +++ b/mk/rte.sdkconfig.mk
> @@ -70,6 +70,11 @@ config: notemplate
>  else
>  config: $(RTE_OUTPUT)/include/rte_config.h $(RTE_OUTPUT)/Makefile
Not sure if this should go under this rule, or "$(RTE_OUTPUT)/.config:"
and should work with ".config_tmp".

>  	$(Q)$(MAKE) depdirs
> +	tac $(RTE_OUTPUT)/.config | awk --field-separator '=' '!/^#/ {print $$1}' | while read config; do \
Why reversing file since already checking all lines one by one in
original file?

And instead of checking each line, it is possible to get list of
duplicates via "sort | uniq -d".

Although less important, file comments also tripled in final .config.

> +		if [ $$(grep "^$${config}=" $(RTE_OUTPUT)/.config | wc -l) -gt 1 ]; then \
"grep -c" can be used instead of "grep | wc -l"

> +			sed -i "0,/$${config}/{//d}" $(RTE_OUTPUT)/.config; \
> +		fi; \
> +	done
>  	@echo "Configuration done"
>  endif
>  
>
Christian Ehrhardt June 28, 2016, 4:38 p.m. UTC | #2 
On Tue, Jun 28, 2016 at 6:11 PM, Ferruh Yigit <ferruh.yigit@intel.com>
wrote:

> On 6/13/2016 4:10 PM, Christian Ehrhardt wrote:
> > Due to the hierarchy and the demand to keep the base config shoing all
> > options some options end up multiple times in the .config file.
> >
> > A suggested solution was to filter for duplicates at the end of the
> > actual config step which is implemented here.
> >
> > Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com>
> > ---
> >  mk/rte.sdkconfig.mk | 5 +++++
> >  1 file changed, 5 insertions(+)
> >
> > diff --git a/mk/rte.sdkconfig.mk b/mk/rte.sdkconfig.mk
> > index a3acfe6..734aa06 100644
> > --- a/mk/rte.sdkconfig.mk
> > +++ b/mk/rte.sdkconfig.mk
> > @@ -70,6 +70,11 @@ config: notemplate
> >  else
> >  config: $(RTE_OUTPUT)/include/rte_config.h $(RTE_OUTPUT)/Makefile
> Not sure if this should go under this rule, or "$(RTE_OUTPUT)/.config:"
> and should work with ".config_tmp".
>
> >       $(Q)$(MAKE) depdirs
> > +     tac $(RTE_OUTPUT)/.config | awk --field-separator '=' '!/^#/
> {print $$1}' | while read config; do \
> Why reversing file since already checking all lines one by one in
> original file?
>

Hi,
every other comment is ok I'll rebase and resubmit once I find some time
again.
But for this (tac) the reason is simple - to keep behaviour.
Currently the last one wins.
So if you have
CONFIG_A=n
CONFIG_A=y

Essentially you have
CONFIG_A=y

By the tac and keeping the first occurrence we maintain behavior.
It is interestingly hard to "keep the last occurrence" without such tricks,
but I'm open to suggestions.


> And instead of checking each line, it is possible to get list of
> duplicates via "sort | uniq -d".
>

That would fail for the reasons outlined above.

Although less important, file comments also tripled in final .config.
>
> > +             if [ $$(grep "^$${config}=" $(RTE_OUTPUT)/.config | wc -l)
> -gt 1 ]; then \
> "grep -c" can be used instead of "grep | wc -l"
>
> > +                     sed -i "0,/$${config}/{//d}"
> $(RTE_OUTPUT)/.config; \
> > +             fi; \
> > +     done
> >       @echo "Configuration done"
> >  endif
> >
> >
>
>
Ferruh Yigit June 28, 2016, 4:48 p.m. UTC | #3 
On 6/28/2016 5:38 PM, Christian Ehrhardt wrote:
> On Tue, Jun 28, 2016 at 6:11 PM, Ferruh Yigit <ferruh.yigit@intel.com
> <mailto:ferruh.yigit@intel.com>> wrote:
> 
>     On 6/13/2016 4:10 PM, Christian Ehrhardt wrote:
>     > Due to the hierarchy and the demand to keep the base config shoing all
>     > options some options end up multiple times in the .config file.
>     >
>     > A suggested solution was to filter for duplicates at the end of the
>     > actual config step which is implemented here.
>     >
>     > Signed-off-by: Christian Ehrhardt <christian.ehrhardt@canonical.com
>     <mailto:christian.ehrhardt@canonical.com>>
>     > ---
>     >  mk/rte.sdkconfig.mk <http://rte.sdkconfig.mk> | 5 +++++
>     >  1 file changed, 5 insertions(+)
>     >
>     > diff --git a/mk/rte.sdkconfig.mk <http://rte.sdkconfig.mk> b/mk/rte.sdkconfig.mk
>     <http://rte.sdkconfig.mk>
>     > index a3acfe6..734aa06 100644
>     > --- a/mk/rte.sdkconfig.mk <http://rte.sdkconfig.mk>
>     > +++ b/mk/rte.sdkconfig.mk <http://rte.sdkconfig.mk>
>     > @@ -70,6 +70,11 @@ config: notemplate
>     >  else
>     >  config: $(RTE_OUTPUT)/include/rte_config.h $(RTE_OUTPUT)/Makefile
>     Not sure if this should go under this rule, or "$(RTE_OUTPUT)/.config:"
>     and should work with ".config_tmp".
> 
>     >       $(Q)$(MAKE) depdirs
>     > +     tac $(RTE_OUTPUT)/.config | awk --field-separator '=' '!/^#/ {print $$1}' | while read config; do \
>     Why reversing file since already checking all lines one by one in
>     original file?
> 
> 
> Hi,
> every other comment is ok I'll rebase and resubmit once I find some time
> again.
> But for this (tac) the reason is simple - to keep behaviour.
> Currently the last one wins.

Correct, but if I am not missing something, reversing doesn't help to this,
how lines deleted taking care of this:
sed -i "0,/$${config}/{//d}" $(RTE_OUTPUT)/.config;

sed works on original file, and deletes first occurrence, independent
from lines from bottom to up, or up to bottom fed into it.

> So if you have
> CONFIG_A=n
> CONFIG_A=y
> 
> Essentially you have 
> CONFIG_A=y
> 
> By the tac and keeping the first occurrence we maintain behavior.
> It is interestingly hard to "keep the last occurrence" without such
> tricks, but I'm open to suggestions.
>  
> 
>     And instead of checking each line, it is possible to get list of
>     duplicates via "sort | uniq -d".
> 
>  
> That would fail for the reasons outlined above.
> 
>     Although less important, file comments also tripled in final .config.
> 
>     > +             if [ $$(grep "^$${config}=" $(RTE_OUTPUT)/.config | wc -l) -gt 1 ]; then \
>     "grep -c" can be used instead of "grep | wc -l"
> 
>     > +                     sed -i "0,/$${config}/{//d}"
>     $(RTE_OUTPUT)/.config; \
>     > +             fi; \
>     > +     done
>     >       @echo "Configuration done"
>     >  endif
>     >
>     >
> 
>
Christian Ehrhardt June 30, 2016, 11:57 a.m. UTC | #4 
Hi,
thanks a lot for your feedback.
I looked at it once more and found some issues - the tac was correct, but
at the wrong place.
Also I could simplify the inner section at least a bit.
I agree to the move to the .config target.

I'll submit a v2 now to this thread.
Here is the diff it generates for a:  "make V=1
T=x86_64-native-linuxapp-gcc config"
http://paste.ubuntu.com/18163566/



Christian Ehrhardt
Software Engineer, Ubuntu Server
Canonical Ltd

On Tue, Jun 28, 2016 at 6:48 PM, Ferruh Yigit <ferruh.yigit@intel.com>
wrote:

> On 6/28/2016 5:38 PM, Christian Ehrhardt wrote:
> > On Tue, Jun 28, 2016 at 6:11 PM, Ferruh Yigit <ferruh.yigit@intel.com
> > <mailto:ferruh.yigit@intel.com>> wrote:
> >
> >     On 6/13/2016 4:10 PM, Christian Ehrhardt wrote:
> >     > Due to the hierarchy and the demand to keep the base config shoing
> all
> >     > options some options end up multiple times in the .config file.
> >     >
> >     > A suggested solution was to filter for duplicates at the end of the
> >     > actual config step which is implemented here.
> >     >
> >     > Signed-off-by: Christian Ehrhardt <
> christian.ehrhardt@canonical.com
> >     <mailto:christian.ehrhardt@canonical.com>>
> >     > ---
> >     >  mk/rte.sdkconfig.mk <http://rte.sdkconfig.mk> | 5 +++++
> >     >  1 file changed, 5 insertions(+)
> >     >
> >     > diff --git a/mk/rte.sdkconfig.mk <http://rte.sdkconfig.mk> b/mk/
> rte.sdkconfig.mk
> >     <http://rte.sdkconfig.mk>
> >     > index a3acfe6..734aa06 100644
> >     > --- a/mk/rte.sdkconfig.mk <http://rte.sdkconfig.mk>
> >     > +++ b/mk/rte.sdkconfig.mk <http://rte.sdkconfig.mk>
> >     > @@ -70,6 +70,11 @@ config: notemplate
> >     >  else
> >     >  config: $(RTE_OUTPUT)/include/rte_config.h $(RTE_OUTPUT)/Makefile
> >     Not sure if this should go under this rule, or
> "$(RTE_OUTPUT)/.config:"
> >     and should work with ".config_tmp".
> >
> >     >       $(Q)$(MAKE) depdirs
> >     > +     tac $(RTE_OUTPUT)/.config | awk --field-separator '=' '!/^#/
> {print $$1}' | while read config; do \
> >     Why reversing file since already checking all lines one by one in
> >     original file?
> >
> >
> > Hi,
> > every other comment is ok I'll rebase and resubmit once I find some time
> > again.
> > But for this (tac) the reason is simple - to keep behaviour.
> > Currently the last one wins.
>
> Correct, but if I am not missing something, reversing doesn't help to this,
> how lines deleted taking care of this:
> sed -i "0,/$${config}/{//d}" $(RTE_OUTPUT)/.config;
>
> sed works on original file, and deletes first occurrence, independent
> from lines from bottom to up, or up to bottom fed into it.
>
> > So if you have
> > CONFIG_A=n
> > CONFIG_A=y
> >
> > Essentially you have
> > CONFIG_A=y
> >
> > By the tac and keeping the first occurrence we maintain behavior.
> > It is interestingly hard to "keep the last occurrence" without such
> > tricks, but I'm open to suggestions.
> >
> >
> >     And instead of checking each line, it is possible to get list of
> >     duplicates via "sort | uniq -d".
> >
> >
> > That would fail for the reasons outlined above.
> >
> >     Although less important, file comments also tripled in final .config.
> >
> >     > +             if [ $$(grep "^$${config}=" $(RTE_OUTPUT)/.config |
> wc -l) -gt 1 ]; then \
> >     "grep -c" can be used instead of "grep | wc -l"
> >
> >     > +                     sed -i "0,/$${config}/{//d}"
> >     $(RTE_OUTPUT)/.config; \
> >     > +             fi; \
> >     > +     done
> >     >       @echo "Configuration done"
> >     >  endif
> >     >
> >     >
> >
> >
>
>
diff mbox

Patch

diff --git a/mk/rte.sdkconfig.mk b/mk/rte.sdkconfig.mk
index a3acfe6..734aa06 100644
--- a/mk/rte.sdkconfig.mk
+++ b/mk/rte.sdkconfig.mk
@@ -70,6 +70,11 @@  config: notemplate
 else
 config: $(RTE_OUTPUT)/include/rte_config.h $(RTE_OUTPUT)/Makefile
 	$(Q)$(MAKE) depdirs
+	tac $(RTE_OUTPUT)/.config | awk --field-separator '=' '!/^#/ {print $$1}' | while read config; do \
+		if [ $$(grep "^$${config}=" $(RTE_OUTPUT)/.config | wc -l) -gt 1 ]; then \
+			sed -i "0,/$${config}/{//d}" $(RTE_OUTPUT)/.config; \
+		fi; \
+	done
 	@echo "Configuration done"
 endif