Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cross-project-issues-dev] Enforce Gerrit for Simrel?

Replies to 4 questions or comments below (in chronological the order). Let me know if I have missed any.

= = = = = = =

cross-project-issues-dev-bounces@xxxxxxxxxxx wrote on 01/07/2016 11:03:31 AM:

> From: Mickael Istria <mistria@xxxxxxxxxx>

> To: cross-project-issues-dev@xxxxxxxxxxx,
> Date: 01/07/2016 11:06 AM
> Subject: Re: [cross-project-issues-dev] Enforce Gerrit for Simrel?
> Sent by: cross-project-issues-dev-bounces@xxxxxxxxxxx
>
> ...

>> But I myself would not like to see it *required* to go through
>> "refs/for/master".

> Why so? Have you tried using it?

Of course. I found it to be very useful!

My objection is to "forcing a policy" when there is not a clear and obvious reason to. That is why I was asking for "data" to see if there was a reason to. In a perfect world, the advantage of using /refs/for/master would be so obvious and helpful, and our committers so well educated, there would be no need to "enforce a policy". But again, if there was data to the contrary, I could be easily convinced.

As for your other questions, I am not sure I understand them. But for "history of failures" what is saved in Hudson is all there is. As for "which commit" went with it, there is a hash stored with each build (look for 'createBuildinfo' in the log) but that is for the "whole repo" that is built ... I think you would have to work backwards from there to find the last several commits (any of which might be the "reason" for the failure). I hope that is close to answering your questions. It might be easier to "monitor builds" from here on, for a month or two to collect data supporting "requiring it" vs. "encouraging it".

= = = = = = =

cross-project-issues-dev-bounces@xxxxxxxxxxx wrote on 01/07/2016 12:08:21 PM:

> From: Konstantin Komissarchik <konstantin.komissarchik@xxxxxxxxxx>

> To: Mickael Istria <mistria@xxxxxxxxxx>, "cross-project-issues-
> dev@xxxxxxxxxxx" <cross-project-issues-dev@xxxxxxxxxxx>,

> Date: 01/07/2016 12:08 PM
> Subject: Re: [cross-project-issues-dev] Enforce Gerrit for Simrel?
> Sent by: cross-project-issues-dev-bounces@xxxxxxxxxxx
>
> -1 to forcing the use of refs/for/x

>  
> The validation results are unreliable as the aggregation build is
> not safely reproducible. A failed validation result is just as
> likely to be due to someone changing their already-contributed repository.

>  

In my experience, this is very rare in the Gerrit-triggered Hudson builds. It is, after all, building "only your change".

Granted, there is a small window where someone might press "commit" just a minute or two before your job runs. But seems to occur much less frequently -- by an order of magnitude -- than before, when many projects would "break HEAD" (and leave it broken for a day or more). Plus, if you are sure that is the case (i.e. Hudson build is "erroneous" and your contribution is OK) then you can override Hudson's -1. I think it takes another reviewer, or something similar.  


= = = = = =

cross-project-issues-dev-bounces@xxxxxxxxxxx wrote on 01/07/2016 12:30:21 PM:

> From: Mickael Istria <mistria@xxxxxxxxxx>

> To: cross-project-issues-dev@xxxxxxxxxxx,
> Date: 01/07/2016 12:30 PM
> Subject: Re: [cross-project-issues-dev] Enforce Gerrit for Simrel?
> Sent by: cross-project-issues-dev-bounces@xxxxxxxxxxx
>
> ...

> Good. I wasn't aware of that. That's a major step forward.
> However a quick lookup on b3aggrcon files shows that many (probably
> the majority) of contributions seems to ignore that rule; as they
> either don't specify a versionRange or specify a versionRange in a
> [a.b.c,x.y.z] way.
>

Yep. You have stated your desire to "test for that" ... so have at it.
(But, do read the "rule" carefully ... it is an either/or type. One condition is easy to test for with simply the text input. The other would have to look at the source repository and make sure it was a 'simple' repository (not a composite).

= = = = = =

cross-project-issues-dev-bounces@xxxxxxxxxxx wrote on 01/07/2016 12:37:01 PM:

> From: Konstantin Komissarchik <konstantin.komissarchik@xxxxxxxxxx>

> To: Doug Schaefer <cdtdoug@xxxxxxxxx>, Cross project issues <cross-
> project-issues-dev@xxxxxxxxxxx>,

> Date: 01/07/2016 12:37 PM
> Subject: Re: [cross-project-issues-dev] Enforce Gerrit for Simrel?
> Sent by: cross-project-issues-dev-bounces@xxxxxxxxxxx
>
> Interesting. When did this happen? There are many contributions
> still using mutable URLs.

>  

The requirement was decided and stated in mid-december:

https://dev.eclipse.org/mhonarc/lists/cross-project-issues-dev/msg12804.html

But read it carefully. I do not think it is exactly what you wanted -- but an easy step to take us 80% in the right direction.

Let the education begin!




Back to the top