On (29/08/16 07:09), Fabiano Fidêncio wrote:
Hoiwdy!
On Fri, Aug 19, 2016 at 1:08 AM, Fabiano Fidêncio <fidencio(a)redhat.com> wrote:
> This patch is a first attempt to make "make intgcheck" less
> painful/time consuming than it is now.
>
> Although the patch provides a good improvement on having 5 new
> targets, I know it's still not ideal. The ideal case, IMO, would be
> being able to select which subset of tests would be run, but that's an
> improvement that can be done later on.
>
> Tips are welcome.
>
> Best Regards,
After a few dicussions and some tests done when we met personally, I
found out that the patch was broken.
Taking Lukaš suggestion (almost) I've merged the -prepare and
-configure parts and re-worked the way we get the prefix.
Now it seems to be working!
Best Regards,
--
Fabiano Fidêncio
From a33e1dffd063845e709f5fc1cfec93c330445ab4 Mon Sep 17 00:00:00
2001
From: =?UTF-8?q?Fabiano=20Fid=C3=AAncio?= <fidencio(a)redhat.com>
Date: Thu, 18 Aug 2016 16:24:17 +0200
Subject: [PATCH v2] BUILD: Add a few more targets for intg tests
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Running "make intgcheck" has been proven to be a bit painful (mainly
when the developer is just writing down a single test case), as it
cleans up the build directory and fireis a new build before, finally,
run the tests.
In order to make it a little less painful, let's break the whole
operation into 4 new targets:
intgcheck-{prepare,build,run,clean}.
As expected, "make intgcheck" calls these 4 new operations in the same
order they were presented, not changing then the current behavior.
Each operation will trigger the previous one in case there is no
"$$prefix" directory created and the directory is _only_ created in the
very first operation (intghceck-prepare).
Signed-off-by: Fabiano Fidêncio <fidencio(a)redhat.com>
---
Makefile.am | 52 ++++++++++++++++++++++++++++++++++++----------------
1 file changed, 36 insertions(+), 16 deletions(-)
diff --git a/Makefile.am b/Makefile.am
index 30d874e..8372b92 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -3076,30 +3076,50 @@ endif
# Integration tests #
#####################
-intgcheck:
+intgcheck-prepare:
echo "temporarily disabled"
set -e; \
- rm -Rf intg; \
- $(MKDIR_P) intg/bld; \
- : Use /hopefully/ short prefix to keep D-Bus socket path short; \
- prefix=`mktemp --tmpdir --directory sssd-intg.XXXXXXXX`; \
- $(LN_S) "$$prefix" intg/pfx; \
- cd intg/bld; \
+ rm -Rf intg ; \
+ $(MKDIR_P) intg/bld ; \
+ : Use /hopefully/ short prefix to keep D-Bus socket path short ; \
+ prefix=`mktemp --tmpdir --directory sssd-intg.XXXXXXXX` ; \
+ $(LN_S) "$$prefix" intg/pfx ; \
+ cd intg/bld ; \
$(abs_top_srcdir)/configure \
- --prefix="$$prefix" \
+ --prefix=$$prefix \
--with-ldb-lib-dir="$$prefix"/lib/ldb \
--enable-intgcheck-reqs \
--without-semanage \
- $(INTGCHECK_CONFIGURE_FLAGS); \
- $(MAKE) $(AM_MAKEFLAGS); \
- : Force single-thread install to workaround concurrency issues; \
- $(MAKE) $(AM_MAKEFLAGS) -j1 install; \
- : Remove .la files from LDB module directory to avoid loader warnings; \
- rm "$$prefix"/lib/ldb/*.la; \
- $(MAKE) $(AM_MAKEFLAGS) -C src/tests/intg intgcheck-installed; \
- cd ../..; \
+ $(INTGCHECK_CONFIGURE_FLAGS) ; \
+ cd ../..
+
+intgcheck-build:
+ if [ ! -d intg/pfx ]; then $(MAKE) intgcheck-prepare; fi ; \
+ prefix=`readlink -e intg/pfx` ; \
+ cd intg/bld ; \
+ $(MAKE) $(AM_MAKEFLAGS) ; \
+ : Force single-thread install to workaround concurrency issues ; \
+ $(MAKE) $(AM_MAKEFLAGS) -j1 install ; \
+ : Remove .la files from LDB module directory to avoid loader warnings ; \
+ rm "$$prefix"/lib/ldb/*.la ; \
+ cd ../..
+
+intgcheck-run:
+ if [ ! -d intg/pfx ]; then $(MAKE) intgcheck-build; fi ; \
+ cd intg/bld ; \
+ $(MAKE) $(AM_MAKEFLAGS) -C src/tests/intg intgcheck-installed ; \
+ cd ../..
+
+intgcheck-clean:
+ prefix=`readlink -e intg/pfx` ; \
rm -Rf "$$prefix" intg
+intgcheck:
+ $(MAKE) intgcheck-prepare
+ $(MAKE) intgcheck-build
I would merge intgcheck-prepare and intgcheck-build bas
well.
The reason is quite simple it simplify *my* workflow of
test development.
My current workflow is
* run make intgcheck
* break it before execution of pytest
* change dir to "intg/build"
* finish preparation
make install
* run tests (or just a particular test)
make -C src/tests/intg intgcheck-installed INTGCHECK_PYTEST_ARGS="-k
test_netgroup.py"
(optional)
* clean everything in custom prefix (after breaking stuff there)
rm -rf ../pfx/*
* prepare stuff for test one more time.
make install
* run tests one more time
make -C src/tests/intg intgcheck-installed INTGCHECK_PYTEST_ARGS="-k
test_netgroup.py"
And now the question is: Do others want to use the same workflow
or we should prefer to run all commands from current
build_dir and not from special build_dir for integration tests.
LS