* ceph-disk and 'make' @ 2016-09-19 0:40 Sage Weil 2016-09-19 5:54 ` Loic Dachary 0 siblings, 1 reply; 9+ messages in thread From: Sage Weil @ 2016-09-19 0:40 UTC (permalink / raw) To: loic; +Cc: ceph-devel I noticed that with the new cmake when I do a 'make' the ceph-disk target isn't built. And ctest will fail most of its unit tests. It looks like it is part of the 'tests' target, though. Is it deliberate that a simple 'make' doesn't build everything? Thanks! sage ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: ceph-disk and 'make' 2016-09-19 0:40 ceph-disk and 'make' Sage Weil @ 2016-09-19 5:54 ` Loic Dachary 2016-09-19 7:35 ` Loic Dachary 0 siblings, 1 reply; 9+ messages in thread From: Loic Dachary @ 2016-09-19 5:54 UTC (permalink / raw) To: Sage Weil; +Cc: ceph-devel On 19/09/2016 02:40, Sage Weil wrote: > I noticed that with the new cmake when I do a 'make' the ceph-disk target > isn't built. And ctest will fail most of its unit tests. It looks like > it is part of the 'tests' target, though. Is it deliberate that a simple > 'make' doesn't build everything? I think make should build it. I'll look into it. Cheers -- Loïc Dachary, Artisan Logiciel Libre ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: ceph-disk and 'make' 2016-09-19 5:54 ` Loic Dachary @ 2016-09-19 7:35 ` Loic Dachary 2016-09-19 18:30 ` Gregory Farnum 0 siblings, 1 reply; 9+ messages in thread From: Loic Dachary @ 2016-09-19 7:35 UTC (permalink / raw) To: Sage Weil; +Cc: ceph-devel After running ./install-deps.sh mkdir build cd build cmake ../ make I have loic@a64e7ff3b216:~/ceph-ubuntu-14.04-loic/build$ ls -l src/ceph-disk ls -l src/ceph-disk total 28 drwxr-xr-x 4 loic loic 4096 Sep 19 05:57 CMakeFiles -rw-r--r-- 1 loic loic 313 Sep 19 05:57 CTestTestfile.cmake -rw-r--r-- 1 loic loic 7240 Sep 19 05:57 Makefile lrwxrwxrwx 1 loic loic 58 Sep 19 06:00 README.rst -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/README.rst lrwxrwxrwx 1 loic loic 57 Sep 19 06:00 ceph_disk -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/ceph_disk -rw-r--r-- 1 loic loic 1878 Sep 19 05:57 cmake_install.cmake lrwxrwxrwx 1 loic loic 64 Sep 19 06:00 requirements.txt -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/requirements.txt lrwxrwxrwx 1 loic loic 56 Sep 19 06:00 setup.py -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/setup.py lrwxrwxrwx 1 loic loic 69 Sep 19 06:00 test-requirements.txt -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/test-requirements.txt which I interpret as make building ceph-disk. Did you expect something else ? Cheers On 19/09/2016 07:54, Loic Dachary wrote: > > > On 19/09/2016 02:40, Sage Weil wrote: >> I noticed that with the new cmake when I do a 'make' the ceph-disk target >> isn't built. And ctest will fail most of its unit tests. It looks like >> it is part of the 'tests' target, though. Is it deliberate that a simple >> 'make' doesn't build everything? > > I think make should build it. I'll look into it. > > Cheers > -- Loïc Dachary, Artisan Logiciel Libre ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: ceph-disk and 'make' 2016-09-19 7:35 ` Loic Dachary @ 2016-09-19 18:30 ` Gregory Farnum 2016-09-20 15:28 ` kefu chai 2016-09-20 22:14 ` Dan Mick 0 siblings, 2 replies; 9+ messages in thread From: Gregory Farnum @ 2016-09-19 18:30 UTC (permalink / raw) To: Loic Dachary; +Cc: Sage Weil, ceph-devel On Sun, Sep 18, 2016 at 5:40 PM, Sage Weil <sweil@redhat.com> wrote: > I noticed that with the new cmake when I do a 'make' the ceph-disk target > isn't built. And ctest will fail most of its unit tests. It looks like > it is part of the 'tests' target, though. Is it deliberate that a simple > 'make' doesn't build everything? Dunno about this specific case, but in general we don't want *everything* to be built with a simple "make", since it slows down compile (and, especially, linking). I don't think ceph-disk is used during development for anything except the tests so this would make some sense, right? On Mon, Sep 19, 2016 at 12:35 AM, Loic Dachary <loic@dachary.org> wrote: > After running > > ./install-deps.sh > mkdir build > cd build > cmake ../ > make > > I have > > loic@a64e7ff3b216:~/ceph-ubuntu-14.04-loic/build$ ls -l src/ceph-disk > ls -l src/ceph-disk > total 28 > drwxr-xr-x 4 loic loic 4096 Sep 19 05:57 CMakeFiles > -rw-r--r-- 1 loic loic 313 Sep 19 05:57 CTestTestfile.cmake > -rw-r--r-- 1 loic loic 7240 Sep 19 05:57 Makefile > lrwxrwxrwx 1 loic loic 58 Sep 19 06:00 README.rst -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/README.rst > lrwxrwxrwx 1 loic loic 57 Sep 19 06:00 ceph_disk -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/ceph_disk > -rw-r--r-- 1 loic loic 1878 Sep 19 05:57 cmake_install.cmake > lrwxrwxrwx 1 loic loic 64 Sep 19 06:00 requirements.txt -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/requirements.txt > lrwxrwxrwx 1 loic loic 56 Sep 19 06:00 setup.py -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/setup.py > lrwxrwxrwx 1 loic loic 69 Sep 19 06:00 test-requirements.txt -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/test-requirements.txt > > which I interpret as make building ceph-disk. Did you expect something else ? Shouldn't any "built" stuff in cmake be going in the build directory rather than the ceph/src one? -Greg ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: ceph-disk and 'make' 2016-09-19 18:30 ` Gregory Farnum @ 2016-09-20 15:28 ` kefu chai 2016-09-20 18:04 ` Ken Dreyer 2016-09-20 22:14 ` Dan Mick 1 sibling, 1 reply; 9+ messages in thread From: kefu chai @ 2016-09-20 15:28 UTC (permalink / raw) To: Gregory Farnum, Marcus Watts; +Cc: Loic Dachary, Sage Weil, ceph-devel On Tue, Sep 20, 2016 at 2:30 AM, Gregory Farnum <gfarnum@redhat.com> wrote: > On Sun, Sep 18, 2016 at 5:40 PM, Sage Weil <sweil@redhat.com> wrote: >> I noticed that with the new cmake when I do a 'make' the ceph-disk target >> isn't built. And ctest will fail most of its unit tests. It looks like >> it is part of the 'tests' target, though. Is it deliberate that a simple >> 'make' doesn't build everything? Sage, it is on purpose. "make" or "make all" should only build the artifacts supposed to be installed. this was pointed out by Marcus. and i am inclined to agree with him on this. > > Dunno about this specific case, but in general we don't want > *everything* to be built with a simple "make", since it slows down > compile (and, especially, linking). I don't think ceph-disk is used > during development for anything except the tests so this would make > some sense, right? agreed, actually, a test-ready ceph-disk needs venv which pip to install the dependencies, but debian pbuilder uses a cleanroom for building packages in which no network access is allowed by default. so if we have builded ceph-disk while "make", pbuilder will have to enable the network access from within the cleanroom. there is a bug filed against jewel on this [1] because we always have ceph-disk venv as a dependency of "make all". -- [1] http://tracker.ceph.com/issues/16169 > > On Mon, Sep 19, 2016 at 12:35 AM, Loic Dachary <loic@dachary.org> wrote: >> After running >> >> ./install-deps.sh >> mkdir build >> cd build >> cmake ../ >> make >> >> I have >> >> loic@a64e7ff3b216:~/ceph-ubuntu-14.04-loic/build$ ls -l src/ceph-disk >> ls -l src/ceph-disk >> total 28 >> drwxr-xr-x 4 loic loic 4096 Sep 19 05:57 CMakeFiles >> -rw-r--r-- 1 loic loic 313 Sep 19 05:57 CTestTestfile.cmake >> -rw-r--r-- 1 loic loic 7240 Sep 19 05:57 Makefile >> lrwxrwxrwx 1 loic loic 58 Sep 19 06:00 README.rst -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/README.rst >> lrwxrwxrwx 1 loic loic 57 Sep 19 06:00 ceph_disk -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/ceph_disk >> -rw-r--r-- 1 loic loic 1878 Sep 19 05:57 cmake_install.cmake >> lrwxrwxrwx 1 loic loic 64 Sep 19 06:00 requirements.txt -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/requirements.txt >> lrwxrwxrwx 1 loic loic 56 Sep 19 06:00 setup.py -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/setup.py >> lrwxrwxrwx 1 loic loic 69 Sep 19 06:00 test-requirements.txt -> /home/loic/ceph-ubuntu-14.04-loic/src/ceph-disk/test-requirements.txt >> >> which I interpret as make building ceph-disk. Did you expect something else ? > > Shouldn't any "built" stuff in cmake be going in the build directory > rather than the ceph/src one? > -Greg > -- > To unsubscribe from this list: send the line "unsubscribe ceph-devel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html -- Regards Kefu Chai ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: ceph-disk and 'make' 2016-09-20 15:28 ` kefu chai @ 2016-09-20 18:04 ` Ken Dreyer 2016-09-21 2:39 ` kefu chai 0 siblings, 1 reply; 9+ messages in thread From: Ken Dreyer @ 2016-09-20 18:04 UTC (permalink / raw) To: kefu chai Cc: Gregory Farnum, Marcus Watts, Loic Dachary, Sage Weil, ceph-devel On Tue, Sep 20, 2016 at 9:28 AM, kefu chai <tchaikov@gmail.com> wrote: > agreed, actually, a test-ready ceph-disk needs venv which pip to install the > dependencies, but debian pbuilder uses a cleanroom for building packages > in which no network access is allowed by default. I'm curious why we need a virtualenv at all (or why we "pip install" anything) ? The distros should have all the packaged python-* and python3-* dependencies already, right? If we could switch to using the packaged python dependencies for everything, we'd fix the pbuilder USENETWORK issue, and avoid the indeterminate nature of depending on whatever happens to be going on with pypi. Maybe I'm missing something, though? - Ken ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: ceph-disk and 'make' 2016-09-20 18:04 ` Ken Dreyer @ 2016-09-21 2:39 ` kefu chai 2016-09-21 5:07 ` Loic Dachary 0 siblings, 1 reply; 9+ messages in thread From: kefu chai @ 2016-09-21 2:39 UTC (permalink / raw) To: Ken Dreyer Cc: Gregory Farnum, Marcus Watts, Loic Dachary, Sage Weil, ceph-devel On Wed, Sep 21, 2016 at 2:04 AM, Ken Dreyer <kdreyer@redhat.com> wrote: > On Tue, Sep 20, 2016 at 9:28 AM, kefu chai <tchaikov@gmail.com> wrote: >> agreed, actually, a test-ready ceph-disk needs venv which pip to install the >> dependencies, but debian pbuilder uses a cleanroom for building packages >> in which no network access is allowed by default. > > I'm curious why we need a virtualenv at all (or why we "pip install" anything) ? > > The distros should have all the packaged python-* and python3-* > dependencies already, right? > > If we could switch to using the packaged python dependencies for > everything, we'd fix the pbuilder USENETWORK issue, and avoid the > indeterminate nature of depending on whatever happens to be going on > with pypi. Maybe I'm missing something, though? we could. the python modules installed by pip are only used for testing purpose, like tox, pytest, flake8, testtools, fixtures, i think we can add them to the Build-Depends list. and let pbuilder to take care of them. @loic what do you think? > > - Ken -- Regards Kefu Chai ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: ceph-disk and 'make' 2016-09-21 2:39 ` kefu chai @ 2016-09-21 5:07 ` Loic Dachary 0 siblings, 0 replies; 9+ messages in thread From: Loic Dachary @ 2016-09-21 5:07 UTC (permalink / raw) To: kefu chai, Ken Dreyer; +Cc: Gregory Farnum, Marcus Watts, Sage Weil, ceph-devel On 21/09/2016 04:39, kefu chai wrote: > On Wed, Sep 21, 2016 at 2:04 AM, Ken Dreyer <kdreyer@redhat.com> wrote: >> On Tue, Sep 20, 2016 at 9:28 AM, kefu chai <tchaikov@gmail.com> wrote: >>> agreed, actually, a test-ready ceph-disk needs venv which pip to install the >>> dependencies, but debian pbuilder uses a cleanroom for building packages >>> in which no network access is allowed by default. >> >> I'm curious why we need a virtualenv at all (or why we "pip install" anything) ? >> >> The distros should have all the packaged python-* and python3-* >> dependencies already, right? >> >> If we could switch to using the packaged python dependencies for >> everything, we'd fix the pbuilder USENETWORK issue, and avoid the >> indeterminate nature of depending on whatever happens to be going on >> with pypi. Maybe I'm missing something, though? > > we could. the python modules installed by pip are only used for > testing purpose, like tox, pytest, flake8, testtools, fixtures, i > think we can add them to the Build-Depends list. and let pbuilder to > take care of them. > > @loic what do you think? It's worth a shot. The primary reason for encapsulating this into install-deps.sh was to cope with all supported operating systems with just two package file / directories (I.e. one debian directory with no #if/#else capability and one spec file). More often than not satisfying python requirements with packages is challenging, to say the least (even pip). That being said, the older operating systems are no longer supported and this may not be an issue right now. And when/if it becomes an issue once again in the future we can probably deal with it in a better way. Cheers -- Loïc Dachary, Artisan Logiciel Libre ^ permalink raw reply [flat|nested] 9+ messages in thread
* Re: ceph-disk and 'make' 2016-09-19 18:30 ` Gregory Farnum 2016-09-20 15:28 ` kefu chai @ 2016-09-20 22:14 ` Dan Mick 1 sibling, 0 replies; 9+ messages in thread From: Dan Mick @ 2016-09-20 22:14 UTC (permalink / raw) To: Gregory Farnum, Loic Dachary; +Cc: Sage Weil, ceph-devel On 09/19/2016 11:30 AM, Gregory Farnum wrote: > On Sun, Sep 18, 2016 at 5:40 PM, Sage Weil <sweil@redhat.com> wrote: >> I noticed that with the new cmake when I do a 'make' the ceph-disk target >> isn't built. And ctest will fail most of its unit tests. It looks like >> it is part of the 'tests' target, though. Is it deliberate that a simple >> 'make' doesn't build everything? > > Dunno about this specific case, but in general we don't want > *everything* to be built with a simple "make", since it slows down > compile (and, especially, linking). I don't think ceph-disk is used > during development for anything except the tests so this would make > some sense, right? I think I disagree; make -> make all-> make everything is certainly my assumption. There's no reason not to have other targets that build less, but I expect "make" to default to "build everything that's configured". (That doesn't include tests, and doesn't install anything, but if it *would be* installed by make install, I expect it to be constructed.) ^ permalink raw reply [flat|nested] 9+ messages in thread
end of thread, other threads:[~2016-09-21 5:07 UTC | newest] Thread overview: 9+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2016-09-19 0:40 ceph-disk and 'make' Sage Weil 2016-09-19 5:54 ` Loic Dachary 2016-09-19 7:35 ` Loic Dachary 2016-09-19 18:30 ` Gregory Farnum 2016-09-20 15:28 ` kefu chai 2016-09-20 18:04 ` Ken Dreyer 2016-09-21 2:39 ` kefu chai 2016-09-21 5:07 ` Loic Dachary 2016-09-20 22:14 ` Dan Mick
This is an external index of several public inboxes, see mirroring instructions on how to clone and mirror all data and code used by this external index.