On Mon, Sep 19, 2016 at 6:49 PM, Vojtech Szocs <vszocs@redhat.com> wrote:
----- Original Message -----
From: "Sandro Bonazzola" <sbonazzo@redhat.com> To: "Eyal Edri" <eedri@redhat.com> Cc: "Juan Hernandez" <jhernand@redhat.com>, "infra" <infra@ovirt.org>, "devel" <devel@ovirt.org> Sent: Monday, September 19, 2016 8:41:12 AM Subject: Re: [ovirt-devel] Dropping rpm build from ovirt-engine check-merged.sh
On Sun, Sep 18, 2016 at 4:18 PM, Eyal Edri < eedri@redhat.com > wrote:
Hi,
Following [1] I'd like to propose to remove rpm building from the 'check-merged.sh' script from ovirt-engine (master for now).
The job [2] takes on avg 15 min while actually the rpms are built already in check-patch (with gwt draft mode if needed) and runs exactly the same build rpm command as check-patch [3].
So there isn't real value in running exactly the same rpm build post merge, and we already build full permutation mode in 'build-artifacts.sh'.
Any reason to keep it? We can cut down valuable time in CI if we drop it and vacant more time for more meaningful tests.
This depends on the flow: if we make check_merge gating to the merge and to the build we should keep the rpm build becuase at merge a rebase is done automatically. If there's not gating process performed by check-merge then I agree in dropping rpm build.
First of all, for oVirt Engine snapshot (non-release) builds, we should avoid doing a full GWT compilation [all browsers x all locales]. That's simply because the full GWT compilation is too costly for CI to run on a regular basis.
+1, It also makes us find regressions slower because it takes more than 1 hour to build + consume very valuable resource which is BM slave.
Doing [Firefox + Chrome x English = 2 permutations] && [draft GWT build option enabled] for snapshot builds should be enough.
The *only* downside of not doing a full GWT build is that problems with localization resources [e.g. non-English .properties files] will not be reported by the GWT compiler. But we have Java unit tests to cover this, so it should be OK. (And if not, we must improve those tests.)
In general, we should do a full GWT compilation only for release builds. Eyal mentioned at [1] that his team is working on
'build official manual' option to standard CI
Already merged and working for most projects.who requested it, also part of standard CI.
so that would be a great place to document & perform the full GWT build.
As for build jobs: if `check-merged` is indeed acting as gate for merge [fail of `check-merged` => patch not merged
its not the case today, we'll need to install Zuul for that, which is in the plan (including system tests).
, `build-artifacts` does not execute], then it actually makes sense to:
- keep `check-merged` doing a (draft) GWT build + Engine RPM build - maybe skip GWT build in `check-patch` [right now, there's detection logic => if frontend files were changed, do GWT build]
Otherwise, if `check-merged` doesn't act as gate for merge, we can just skip GWT build / Engine RPM build for that script.
Yea, doesn't act as gate, it just run post merge.
[1] https://ovirt-jira.atlassian.net/browse/OVIRT-416 [2] http://jenkins.ovirt.org/job/ovirt-engine_master_check-
merged-el7-x86_64/buildTimeTrend
[3] rpmbuild \ -D "_rpmdir $PWD/output" \ -D "_topmdir $PWD/rpmbuild" \ -D "release_suffix ${SUFFIX}" \ -D "ovirt_build_ut $BUILD_UT" \ -D "ovirt_build_extra_flags $EXTRA_BUILD_FLAGS" \ -D "ovirt_build_draft 1" \ --rebuild output/*.src.rpm
-- Eyal Edri Associate Manager RHV DevOps EMEA ENG Virtualization R&D Red Hat Israel
phone: +972-9-7692018 irc: eedri (on #tlv #rhev-dev #rhev-integ)
-- Sandro Bonazzola Better technology. Faster innovation. Powered by community collaboration. See how it works at redhat.com
_______________________________________________ Devel mailing list Devel@ovirt.org http://lists.ovirt.org/mailman/listinfo/devel
-- Eyal Edri Associate Manager RHV DevOps EMEA ENG Virtualization R&D Red Hat Israel phone: +972-9-7692018 irc: eedri (on #tlv #rhev-dev #rhev-integ)