Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
[platform-releng-dev] More on "when we fail the build" ..

In my never ending quest to be honest, and open, my initial response to Dani's question was not quite right.

By chance, I noticed our invocation was not what I remembered ... I thought we used --fail-at-end, but it is currently set to --fail-fast. There's a possibility --fail-never might still produce packages if a test bundle failed to compile. But, also possible the build would still fail later -- in a more confusing way -- or we would have to be improve the build in some way, to continue under "failure" circumstances.

so have opened a bug to investigate:

Bug 477184- Investigate effects of different failure options

I'll try to track down the history of when and why changed to "fail fast" and at least change back to "fail at end" (which will still fail after all bundles compiled, I believe) and with less priority, investigate effects of --fail-never.

Thanks,






From:        David M Williams/Raleigh/IBM@IBMUS
To:        "Eclipse platform release engineering list." <platform-releng-dev@xxxxxxxxxxx>,
Date:        09/10/2015 05:46 AM
Subject:        Re: [platform-releng-dev] 4.6.0 N-Build: N20150908-2000 -BUILDFAILED
Sent by:        platform-releng-dev-bounces@xxxxxxxxxxx





> From: Daniel Megert <daniel_megert@xxxxxxxxxx>
> To: "Eclipse platform release engineering list." <platform-releng-
> dev@xxxxxxxxxxx>,
> Date: 09/09/2015 02:54 AM
> Subject: Re: [platform-releng-dev] 4.6.0 N-Build: N20150908-2000 -
> BUILD FAILED
> Sent by: platform-releng-dev-bounces@xxxxxxxxxxx
>
> > [ERROR] Failed to execute goal org.eclipse.tycho:tycho-compiler-
> plugin:0.23.1:compile (default-compile) on project
> org.eclipse.core.tests.resources: Compilation failure: Compilation failure:
>
> Why does the entire build fail because of a compile error in a test
> bundle? I thought we publish those builds that have no errors.


In short, we continue "as much as we reasonably can". If Tycho had it's (default) way,
it would stop at the first compile error ... we at least continue compiling in case there's
additional ones to know about.


I believe the PDE build could "continue further" since it's builds were feature based ...
easy to distinguish runtime code from test code. Plus, I, as I recall, I think PDE would still
"assemble" the bundle, even if it had compile errors, whereas Tycho will not.


Since Tycho is Git repository based, there is no easy way to tell if test or runtime code caused the failure.
And, even if we could do some "custom coding" to get around that, I am not sure how far we
could get with it before Tycho would fail again due to it's own ideas of what
it should build, and what it should not. (For example, it would at least fail at creating
p2 repositories, because something would be missing.)


Hope that helps. If anyone knows something I don't, let me know.


_______________________________________________
platform-releng-dev mailing list
platform-releng-dev@xxxxxxxxxxx
To change your delivery options, retrieve your password, or unsubscribe from this list, visit
https://dev.eclipse.org/mailman/listinfo/platform-releng-dev


Back to the top