[dpdk-dev,3/4] mk: add new test-run make rule

  Since "make test" and "make test-build" does dependency resolving, they
check for all dependent components (lib and drivers) which takes a few
seconds.

This is a good feature during development, but if the target is just
running unit test, that step is unnecessary, it is possible to compile
onece and run unit test multiple times, without checking any code update

For this purpose, a new make rule "make test-run" added. Which just runs
the unit test, expects that unit test already compiled.

Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
---
 mk/rte.sdkroot.mk | 4 ++--
 mk/rte.sdktest.mk | 4 ++--
 2 files changed, 4 insertions(+), 4 deletions(-)
  

On Wed, Feb 15, 2017 at 03:26:31PM +0000, Ferruh Yigit wrote:
> Since "make test" and "make test-build" does dependency resolving, they
> check for all dependent components (lib and drivers) which takes a few
> seconds.
> 
> This is a good feature during development, but if the target is just
> running unit test, that step is unnecessary, it is possible to compile
> onece and run unit test multiple times, without checking any code update
> 
> For this purpose, a new make rule "make test-run" added. Which just runs
> the unit test, expects that unit test already compiled.
> 
> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> ---
Sorry, I just don't see the point of having an extra command to maintain
and document for the sake of a few seconds on a unit test run. How long
does the run itself take compared to the time to check dependencies? 

/Bruce
  

On 2/15/2017 5:07 PM, Bruce Richardson wrote:
> On Wed, Feb 15, 2017 at 03:26:31PM +0000, Ferruh Yigit wrote:
>> Since "make test" and "make test-build" does dependency resolving, they
>> check for all dependent components (lib and drivers) which takes a few
>> seconds.
>>
>> This is a good feature during development, but if the target is just
>> running unit test, that step is unnecessary, it is possible to compile
>> onece and run unit test multiple times, without checking any code update
>>
>> For this purpose, a new make rule "make test-run" added. Which just runs
>> the unit test, expects that unit test already compiled.
>>
>> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
>> ---
> Sorry, I just don't see the point of having an extra command to maintain
> and document for the sake of a few seconds on a unit test run. How long
> does the run itself take compared to the time to check dependencies? 

It is matter of choice, it does not take much time for "make test", but
still I thought it is handy to have a rule only to run the test.

I don't expect much maintenance cost with this, also I don't have strong
opinion to keep it.

> 
> /Bruce
>
  

On Wed, Feb 15, 2017 at 05:41:08PM +0000, Ferruh Yigit wrote:
> On 2/15/2017 5:07 PM, Bruce Richardson wrote:
> > On Wed, Feb 15, 2017 at 03:26:31PM +0000, Ferruh Yigit wrote:
> >> Since "make test" and "make test-build" does dependency resolving, they
> >> check for all dependent components (lib and drivers) which takes a few
> >> seconds.
> >>
> >> This is a good feature during development, but if the target is just
> >> running unit test, that step is unnecessary, it is possible to compile
> >> onece and run unit test multiple times, without checking any code update
> >>
> >> For this purpose, a new make rule "make test-run" added. Which just runs
> >> the unit test, expects that unit test already compiled.
> >>
> >> Signed-off-by: Ferruh Yigit <ferruh.yigit@intel.com>
> >> ---
> > Sorry, I just don't see the point of having an extra command to maintain
> > and document for the sake of a few seconds on a unit test run. How long
> > does the run itself take compared to the time to check dependencies? 
> 
> It is matter of choice, it does not take much time for "make test", but
> still I thought it is handy to have a rule only to run the test.
> 
> I don't expect much maintenance cost with this, also I don't have strong
> opinion to keep it.
> 
Yes, I suppose it isn't much maintenance cost indeed, so if nobody else
cares, I'm ok with it too.

/Bruce
  

2017-02-15 15:26, Ferruh Yigit:
> --- a/mk/rte.sdkroot.mk
> +++ b/mk/rte.sdkroot.mk
> @@ -92,8 +92,8 @@ default: all
>  config showconfigs showversion showversionum:
>  	$(Q)$(MAKE) -f $(RTE_SDK)/mk/rte.sdkconfig.mk $@
>  
> -.PHONY: test fast_test ring_test mempool_test perf_test coverage
> -test fast_test ring_test mempool_test perf_test coverage:
> +.PHONY: test test-run fast_test ring_test mempool_test perf_test coverage
> +test test-run fast_test ring_test mempool_test perf_test coverage:
>  	$(Q)$(MAKE) -f $(RTE_SDK)/mk/rte.sdktest.mk $@

"test" is a shortcut for build + basic tests.
I think test-run can be better named. It runs all the basic tests registered
in autotests. "test-all" would be wrong. What about "test-basic"?
  

On 2/16/2017 9:26 AM, Thomas Monjalon wrote:
> 2017-02-15 15:26, Ferruh Yigit:
>> --- a/mk/rte.sdkroot.mk
>> +++ b/mk/rte.sdkroot.mk
>> @@ -92,8 +92,8 @@ default: all
>>  config showconfigs showversion showversionum:
>>  	$(Q)$(MAKE) -f $(RTE_SDK)/mk/rte.sdkconfig.mk $@
>>  
>> -.PHONY: test fast_test ring_test mempool_test perf_test coverage
>> -test fast_test ring_test mempool_test perf_test coverage:
>> +.PHONY: test test-run fast_test ring_test mempool_test perf_test coverage
>> +test test-run fast_test ring_test mempool_test perf_test coverage:
>>  	$(Q)$(MAKE) -f $(RTE_SDK)/mk/rte.sdktest.mk $@
> 
> "test" is a shortcut for build + basic tests.
> I think test-run can be better named. It runs all the basic tests registered
> in autotests. "test-all" would be wrong. What about "test-basic"?

Right, "test-basic" matches more to existing rules (test-fast,
test-perf, ..), all has hidden "run" action implied, I will use
"test-basic" if there is no objection.

My concerns is "test-basic" "running basic tests without compilation"
may not be obvious for a newcomer.
Should I add a line to "make help" for "test" and "test-*" rules?

Like:
test      Compile tests and run basic unit tests
test-*    Run specific subset of the unit tests

thanks,
ferruh
  

2017-02-16 10:21, Ferruh Yigit:
> On 2/16/2017 9:26 AM, Thomas Monjalon wrote:
> > 2017-02-15 15:26, Ferruh Yigit:
> >> --- a/mk/rte.sdkroot.mk
> >> +++ b/mk/rte.sdkroot.mk
> >> @@ -92,8 +92,8 @@ default: all
> >>  config showconfigs showversion showversionum:
> >>  	$(Q)$(MAKE) -f $(RTE_SDK)/mk/rte.sdkconfig.mk $@
> >>  
> >> -.PHONY: test fast_test ring_test mempool_test perf_test coverage
> >> -test fast_test ring_test mempool_test perf_test coverage:
> >> +.PHONY: test test-run fast_test ring_test mempool_test perf_test coverage
> >> +test test-run fast_test ring_test mempool_test perf_test coverage:
> >>  	$(Q)$(MAKE) -f $(RTE_SDK)/mk/rte.sdktest.mk $@
> > 
> > "test" is a shortcut for build + basic tests.
> > I think test-run can be better named. It runs all the basic tests registered
> > in autotests. "test-all" would be wrong. What about "test-basic"?
> 
> Right, "test-basic" matches more to existing rules (test-fast,
> test-perf, ..), all has hidden "run" action implied, I will use
> "test-basic" if there is no objection.
> 
> My concerns is "test-basic" "running basic tests without compilation"
> may not be obvious for a newcomer.
> Should I add a line to "make help" for "test" and "test-*" rules?
> 
> Like:
> test      Compile tests and run basic unit tests
> test-*    Run specific subset of the unit tests

Yes it looks good.