[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [cross-project-issues-dev] massive amount of unresolved bundles in 2018-09 M3 (JEE package)

Hi,

On Mon, Sep 10, 2018 at 5:59 PM, Ed Willink <ed@xxxxxxxxxxxxx> wrote:
you have your problem, nobody can be sure what you contributed when and some lucky SimRel users may get to see a contribution change mid-build. Worse, if you mess up so that your aggregate repo is bad, the SimRel aggregation is broken until you fix the mess.

IMHO generic contributions should be strongly deprecated.

Documentation makes it clear: https://wiki.eclipse.org/Simrel/Contributing_to_Simrel_Aggregation_Build#Format
However, pragmatically, plugins that do a lot of builds and want them to be continuously integrated don't have a real choice here, unless they're fine to update their aggrcon several times a day... Referencing mutable repos or content is pragmatically more productive, and that's probably the cause of several bad practices.

While generic contributions are pragmatically used by many projects during the "dev" stream, they should indeed be banished while we enter the RC phase.
But as long a we do not automate a failing build for this bad practice, no-one will really spend more care on applying this rule.

What we need to develop is a set of concrete and automatable rules that for a given .aggrcon file can say whether it's valid or not.
Then, we can plug it in the builder so an invalid aggrcon fails the build if we're in RC.