On Mon, Aug 29, 2016 at 10:38:46AM +0200, Lukas Slebodnik wrote:
> On (29/08/16 07:09), Fabiano Fidêncio wrote:
> >Hoiwdy!
> >
> >
> >On Fri, Aug 19, 2016 at 1:08 AM, Fabiano Fidêncio <fiden...@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?= <fiden...@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
> >"$$prefix" directory created and the directory is _only_ created in the
> >very first operation (intghceck-prepare).
> >
> >Signed-off-by: Fabiano Fidêncio <fiden...@redhat.com>
> >---
> > Makefile.am | 52 ++++++++++++++++++++++++++++++++++++----------------
> > 1 file changed, 36 insertions(+), 16 deletions(-)
> >
> >diff --git a/Makefile.am b/Makefile.am
> >index 30d874e..8372b92 100644
> >--- a/Makefile.am
> >+++ b/Makefile.am
> >@@ -3076,30 +3076,50 @@ endif
> > # Integration tests #
> > #####################
> >
> >-intgcheck:
> >+intgcheck-prepare:
> > echo "temporarily disabled"
> > set -e; \
> >- rm -Rf intg; \
> >- $(MKDIR_P) intg/bld; \
> >- : Use /hopefully/ short prefix to keep D-Bus socket path short; \
> >- prefix=`mktemp --tmpdir --directory sssd-intg.XXXXXXXX`; \
> >- $(LN_S) "$$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.
<userstory>
As a developer, I need a way to run a single test without
re-compiling SSSD every time and waiting for all the tests to
finish.
</userstory>
How we get there is not that relevant for me :) I'm fine with compiling
SSSD the first time I run integration tests. It would be nice if
multiple runs after I change SSSD code when debugging would allow me to
only compile the part that changed and run make install, but as a first
iteration, just having a cleaner way to run a single test than running
the full test suite would be nice.
_______________________________________________
sssd-devel mailing list
sssd-devel@lists.fedorahosted.org
https://lists.fedorahosted.org/admin/lists/sssd-devel@lists.fedorahosted.org
