[PATCH] CI: Use make check instead of make-check-wrap

[Lukas Slebodnik]

ehlo,

the 1st patch is prerequisite for the 2nd patch
and it also speed up build on rhel6.

The second patch simplify some magic in CI script.

LS

[Nikolai Kondrashov]

Hi Lukas,

On 03/31/2016 02:41 PM, Lukas Slebodnik wrote:
>  ehlo,
>
>  the 1st patch is prerequisite for the 2nd patch
>  and it also speed up build on rhel6.
>
>  The second patch simplify some magic in CI script. 
The idea is fine with me and is welcome. Thank you, Lukas :)
I have a couple of comments, though.

>  -AM_INIT_AUTOMAKE([-Wall -Wno-portability foreign subdir-objects
> tar-pax])
>  +AM_INIT_AUTOMAKE([-Wall -Wno-portability foreign subdir-objects tar-pax
>  +                  parallel-tests]) 
This is obviously fine.

>        # Not building "tests" due
> tohttps://fedorahosted.org/sssd/ticket/2350
>  -    stage make-tests        make-check-wrap -j $CPU_NUM check -- true
>  +    stage make-tests        make -j $CPU_NUM check 
This changes the behavior to not only build the tests but to also run them.
Is that intentional? If not, then the fix should be adding
"LOG_COMPILER=true".

>       status=0
>       CK_FORK=no \
>           stage make-check-valgrind \
> -                make-check-wrap -j $CPU_NUM check -- \
> -                                libtool --mode=execute \
> -                                    valgrind-condense 99 \
> -                                    "$valgrind_test_pattern" -- \
> -                                    --trace-children=yes \
> -                                    --trace-children-skip='*/bin/*,*/sbin/*'
\
> -                                    --leak-check=full \
> -                                    --gen-suppressions=all \
> -                                    --suppressions="$CI_DIR/sssd.supp" \
> -                                    --verbose ||
> +                make -j $CPU_NUM check -- \
> +                     LOG_COMPILER=libtool \
> +                     LOG_FLAGS="--mode=execute \
> +                                valgrind-condense 99 \
> +                                \"$valgrind_test_pattern\" -- \
> +                                --trace-children=yes \
> +                                --trace-children-skip='*/bin/*,*/sbin/*' \
> +                                --leak-check=full \
> +                                --gen-suppressions=all \
> +                                --suppressions=\"$CI_DIR/sssd.supp\" \
> +                                --verbose" ||
>               status=$?
>       mv "$test_dir" ci-test-dir
>       ((status == 0))
> @@ -279,7 +280,7 @@ function build_coverage()
>
>       # Build everything, including tests
>        # Not building "tests" due
> tohttps://fedorahosted.org/sssd/ticket/2350
>  -    stage make-tests        make-check-wrap -j $CPU_NUM check -- true
>  +    stage make-tests        make -j $CPU_NUM check 
Same here.

Also, if we're removing all the uses of make-check-wrap, then we should also
remove the "make-check-wrap" file itself.

Nick

[Lukas Slebodnik]

On (31/03/16 15:06), Nikolai Kondrashov wrote:
> Hi Lukas,
>
> On 03/31/2016 02:41 PM, Lukas Slebodnik wrote:
> >ehlo,
> >
> >the 1st patch is prerequisite for the 2nd patch
> >and it also speed up build on rhel6.
> >
> >The second patch simplify some magic in CI script.
>
> The idea is fine with me and is welcome. Thank you, Lukas :)
> I have a couple of comments, though.
>
> >-AM_INIT_AUTOMAKE([-Wall -Wno-portability foreign subdir-objects tar-pax])
> >+AM_INIT_AUTOMAKE([-Wall -Wno-portability foreign subdir-objects tar-pax
> >+                  parallel-tests])
>
> This is obviously fine.
>
> >      # Not building "tests" due
> tohttps://fedorahosted.org/sssd/ticket/2350
> >-    stage make-tests        make-check-wrap -j $CPU_NUM check -- true
> >+    stage make-tests        make -j $CPU_NUM check
>
> This changes the behavior to not only build the tests but to also run them.
> Is that intentional? If not, then the fix should be adding
> "LOG_COMPILER=true".
>  
You are right.
I missed that change.

BTW bew version with "LOG_COMPILER=true" is more obvious
than old version "-- true"


> >      status=0
> >      CK_FORK=no \
> >          stage make-check-valgrind \
> >-                make-check-wrap -j $CPU_NUM check -- \
> >-                                libtool --mode=execute \
> >-                                    valgrind-condense 99 \
> >-                                    "$valgrind_test_pattern" -- \
> >-                                    --trace-children=yes \
> >-                                    --trace-children-skip='*/bin/*,*/sbin/*'
> \
> >-                                    --leak-check=full \
> >-                                    --gen-suppressions=all \
> >-                                    --suppressions="$CI_DIR/sssd.supp" \
> >-                                    --verbose ||
> >+                make -j $CPU_NUM check -- \
> >+                     LOG_COMPILER=libtool \
> >+                     LOG_FLAGS="--mode=execute \
> >+                                valgrind-condense 99 \
> >+                                \"$valgrind_test_pattern\" -- \
> >+                                --trace-children=yes \
> >+                                --trace-children-skip='*/bin/*,*/sbin/*' \
> >+                                --leak-check=full \
> >+                                --gen-suppressions=all \
> >+                                --suppressions=\"$CI_DIR/sssd.supp\" \
> >+                                --verbose" ||
> >              status=$?
> >      mv "$test_dir" ci-test-dir
> >      ((status == 0))
> >@@ -279,7 +280,7 @@ function build_coverage()
> >
> >      # Build everything, including tests
> >      # Not building "tests" due
> tohttps://fedorahosted.org/sssd/ticket/2350
> >-    stage make-tests        make-check-wrap -j $CPU_NUM check -- true
> >+    stage make-tests        make -j $CPU_NUM check
>
> Same here.
>  
Done

> Also, if we're removing all the uses of make-check-wrap, then we
> should also
> remove the "make-check-wrap" file itself.
>  
Removed

Updated version is attached.

http://sssd-ci.duckdns.org/logs/job/40/35/summary.html

BTW Feel free to propose better indentation if needed.

LS

[Nikolai Kondrashov]

On 03/31/2016 05:02 PM, Lukas Slebodnik wrote:
>  On (31/03/16 15:06), Nikolai Kondrashov wrote:
>  BTW bew version with "LOG_COMPILER=true" is more obvious
>  than old version "-- true" 
Sure, as long as you know how the value of "LOG_COMPILER" is used :)

>  Updated version is attached.
>
>  http://sssd-ci.duckdns.org/logs/job/40/35/summary.html 
Thanks!

>  BTW Feel free to propose better indentation if needed. 
No, it's fine :)

ACK.

Nick

[Lukas Slebodnik]

On (31/03/16 20:43), Nikolai Kondrashov wrote:
> On 03/31/2016 05:02 PM, Lukas Slebodnik wrote:
> >On (31/03/16 15:06), Nikolai Kondrashov wrote:
> >BTW bew version with "LOG_COMPILER=true" is more obvious
> >than old version "-- true"
>
> Sure, as long as you know how the value of "LOG_COMPILER" is used :)
>
> >Updated version is attached.
> >
> >http://sssd-ci.duckdns.org/logs/job/40/35/summary.html
>
> Thanks!
>
> >BTW Feel free to propose better indentation if needed.
>
> No, it's fine :)
>
> ACK.
>  
master:
* 444a82bd6d68c6f23e05d523ff92d328f6b2ec05
* 5ebdc2391e96cfcc86ebdb8f223e159c00a0d82b

Patches speed up a build on rhel6 therefore pushed
also to stable branch

sssd-1-13:
* e252ba9142933dfacbba177bd07ef7ce3672d4e5
* 5e1e8195304c4f348140c65f41ea843798dfb4e3

LS