Yes,

We're using this migration to slowly drop support for GCC 4 (
https://gem5.atlassian.net/browse/GEM5-218), which i believe is causing
these discrepencies. GCB runs on a Docker Image built with
https://gem5.googlesource.com/public/gem5/+/refs/heads/develop/util/dockerfiles/ubuntu-18.04_all-dependencies/Dockerfile.


I'm currently working on a script that I hope fixes all of Jason's points
so we can fully migrate prior to gem5-20 ( I'm going to try to keep the
name Kokoro because both Jason and I think "Gem5 Cloud Project GCB service
account" is annoyingly long).

--
Dr. Bobby R. Bruce
Room 2235,
Kemper Hall, UC Davis
Davis,
CA, 95616

web: https://www.bobbybruce.net


On Tue, Mar 31, 2020 at 9:27 AM Giacomo Travaglini <
[email protected]> wrote:

> Thanks Jason for the great effort you are putting on this, and for
> updating us on the current state.
>
>
>
> It’s not really relevant to our discussion, but while GCB and kokoro run
> the same tests, they do that under different building environments.
>
> An example is:
>
>
>
> https://gem5-review.googlesource.com/c/public/gem5/+/23527
>
>
>
> Which is failing under kokoro (gcc4.8) while passing under GCB
>
>
>
> Giacomo
>
>
>
>
>
> *From:* Jason Lowe-Power <[email protected]>
> *Sent:* 31 March 2020 16:17
> *To:* Ciro Santilli <[email protected]>
> *Cc:* Giacomo Travaglini <[email protected]>; Nils Asmussen <
> [email protected]>; gem5 Developer List <[email protected]>;
> Bobby Bruce <[email protected]>
> *Subject:* Re: [gem5-dev] Removing kokoro runs
>
>
>
> A couple of things... Bobby can correct me if I'm wrong.
>
>
>
> 1. GCB and kokoro should be running *the same tests*. So they should
> always have the same result. There's no need to wait for kokoro if you have
> a +1 from GCB.
>
> 2. The migration is still in progress. There are a couple of loose ends we
> need to tie up before turning off kokoro.
>
>   - Getting the logs to be publicly available (see below)
>
>   - Requiring a maintainer +1 before running GCB (this is needed to cut
> down on our costs... right now, we're spending between $10 and $60 per day
> depending on the number of changesets.
>
>   - Running in batches instead of every time something is pushed. Right
> now, if you push two versions of the same changeset back to back we test
> both of them... we need to stop doing that. This isn't an "automatic"
> feature of GCB, so we're working on our own implementation of batching.
>
> 3. The permissions are something we're working on. GCB does not allow us
> to make the logs public. Seems like an obvious important feature, but we
> can't find a way. We're working on a system that will automatically dump
> those logs into a publicly available bucket. We'll make sure this done
> before we turn off kokoro.
>
>
>
> At a higher level... it's turned out to be a lot more engineering work to
> get GCB + gerrit working for our use case of CI testing. If anyone else has
> expertise here and would like to contribute to help us get this to
> production quality, we would appreciate it!
>
>
>
> Cheers,
>
> Jason
>
>
>
> On Tue, Mar 31, 2020 at 7:38 AM Ciro Santilli <[email protected]>
> wrote:
>
> Yes, I am logged in.
> ------------------------------
>
> *From:* Giacomo Travaglini <[email protected]>
> *Sent:* Tuesday, March 31, 2020 3:29 PM
> *To:* Nils Asmussen <[email protected]>; gem5 Developer List <
> [email protected]>; Ciro Santilli <[email protected]>; Jason
> Lowe-Power <[email protected]>; Bobby Bruce <[email protected]>
> *Subject:* RE: [gem5-dev] Removing kokoro runs
>
>
>
> Have you tried to login?
>
> Giacomo
>
> -----Original Message-----
> From: Nils Asmussen <[email protected]>
> Sent: 31 March 2020 15:27
> To: gem5 Developer List <[email protected]>; Ciro Santilli <
> [email protected]>; Jason Lowe-Power <[email protected]>; Bobby
> Bruce <[email protected]>; Giacomo Travaglini <[email protected]
> >
> Subject: Re: [gem5-dev] Removing kokoro runs
>
> Hi,
>
> same for me. Google says the permission "cloudbuild.builds.get" is missing.
>
> Best regards,
> Nils
>
>
> On 3/31/20 4:23 PM, Ciro Santilli wrote:
> > Since Giacomo started this thread, I would also want to chime in a
> related question: are you able to see GCB run logs?
> >
> > E.g. at: https://gem5-review.googlesource.com/c/public/gem5/+/27315 GCB
> links to
> https://console.cloud.google.com/cloud-build/builds/bbd716e1-a572-4c69-8c99-612df1755898?project=345032938727&pli=1
> but when I visit that I get: "You do not have sufficient permissions to
> view this page".
> >
> > Do others reproduce/is it expected?
> > ________________________________
> > From: gem5-dev <[email protected]> on behalf of Giacomo
> Travaglini <[email protected]>
> > Sent: Tuesday, March 31, 2020 1:21 PM
> > To: [email protected] <[email protected]>; Jason Lowe-Power <
> [email protected]>; Bobby Bruce <[email protected]>
> > Subject: [gem5-dev] Removing kokoro runs
> >
> > Hi,
> >
> > I have noticed we have started using the Google Cloud service to run our
> pre-submit regressions.
> > At the moment this is happening in parallel with kokoro.
> >
> > Few things to point out:
> >
> >
> >   1.  It is actually possible now to submit a patch without kokoro. GCB
> is faster than kokoro and starts as soon as the patch is pushed regardless
> of the maintainer vote. As soon as you get +2, +1, +1 (from GCB) you can
> submit the patch while kokoro is running (because of the maintainer vote).
> >   2.  I am trying to be diligent and I am waiting for kokoro in any case
> (+2 +1 +1 +1) I might for example break kokoro and prevent a user from
> submitting his own patch
> >
> >
> > Considering this, I'd like to ask: which is the removal plan for kokoro?
> >
> > Kind Regards
> >
> > Giacomo
> >
> >
> >
> >
> > IMPORTANT NOTICE: The contents of this email and any attachments are
> confidential and may also be privileged. If you are not the intended
> recipient, please notify the sender immediately and do not disclose the
> contents to any other person, use it for any purpose, or store or copy the
> information in any medium. Thank you.
> > _______________________________________________
> > gem5-dev mailing list
> > [email protected]
> > http://m5sim.org/mailman/listinfo/gem5-dev
> > _______________________________________________
> > gem5-dev mailing list
> > [email protected]
> > http://m5sim.org/mailman/listinfo/gem5-dev
> >
>
>
> IMPORTANT NOTICE: The contents of this email and any attachments are
> confidential and may also be privileged. If you are not the intended
> recipient, please notify the sender immediately and do not disclose the
> contents to any other person, use it for any purpose, or store or copy the
> information in any medium. Thank you.
>
> IMPORTANT NOTICE: The contents of this email and any attachments are
> confidential and may also be privileged. If you are not the intended
> recipient, please notify the sender immediately and do not disclose the
> contents to any other person, use it for any purpose, or store or copy the
> information in any medium. Thank you.
>
_______________________________________________
gem5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/gem5-dev

Reply via email to