On (29/08/16 12:08), Fabiano Fidêncio wrote:
>On Mon, Aug 29, 2016 at 11:12 AM, Jakub Hrozek <jhro...@redhat.com> wrote:
>> 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.
>
>Can be done. But if that's not a mandatory requirement I still think
>it may be useful for someone else in the way it's split.
>
"prepare" and "build" it's the same for me.
It is a prerequisity for
<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>
>>>
>>> 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"
>>>
>
>That's something really good to now!
>
>>> 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>
>
>I've tried to achieve this in the simplest way possible. Please, take
>a look in the attached patch (that should go on top of the v2 sent
>this morning).
>
>>
>> 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.
Jakub,
as you could see it's possible even with current version :-)
But it's not very user friendly
and "debugging would allow me to only compile the part that changed and run
make install" requires either to prefix all current targets with "intg"
e.g. "intg-all", "intg-check", "intg-install" ...
or to cd into build_dir for integration tests.
>From 4ff76411b6b8230c92345bf73bf88fd76908d0fb Mon Sep 17 00:00:00 2001
>From: =?UTF-8?q?Fabiano=20Fid=C3=AAncio?= <fiden...@redhat.com>
>Date: Mon, 29 Aug 2016 11:59:13 +0200
>Subject: [PATCH] BUILD: Add the possibility to run a single integration test
>
>In order to be able to run a single integration test, let's add one new
>more rule to the Makefile (make intgcheck-run-single-test) that
>receives, from an environment variable called INTGCHECK_SINGLE_TEST, the
>test to be run.
>
>An command line example would be:
>INTGCHECK_SINGLE_TEST="test_netgroup.py" make intgcheck-run-single-test
>---
> Makefile.am | 7 +++++++
> 1 file changed, 7 insertions(+)
>
>diff --git a/Makefile.am b/Makefile.am
>index 8372b92..7e053d7 100644
>--- a/Makefile.am
>+++ b/Makefile.am
>@@ -3114,6 +3114,13 @@ intgcheck-clean:
> prefix=`readlink -e intg/pfx` ; \
> rm -Rf "$$prefix" intg
>
>+intgcheck-run-single-test:
>+ if [ ! -d intg/pfx ]; then $(MAKE) intgcheck-build; fi ; \
>+ cd intg/bld ; \
>+ $(MAKE) $(AM_MAKEFLAGS) -C src/tests/intg intgcheck-installed \
>+ INTGCHECK_PYTEST_ARGS="-k $(INTGCHECK_SINGLE_TEST)" ; \
>+ cd ../..
>+
I'm sorry but it's better to have less targets rather then more targets.
intgcheck-run needs to work with INTGCHECK_PYTEST_ARGS.
IMHO, it's really confusing to have more env variables for the same purpose
LS
_______________________________________________
sssd-devel mailing list
sssd-devel@lists.fedorahosted.org
https://lists.fedorahosted.org/admin/lists/sssd-devel@lists.fedorahosted.org
