|Re: [cross-project-issues-dev] Enforce Gerrit for Simrel?|
On 01/09/2016 08:21 AM, Ed Willink wrote:
IMHO given build-specific repos, the version range is redundant clutter and arguably the source of much of this trouble.The main issue is that, when reading this b3aggrcon file, there is no way to know and test automatically whether the provided repository contents won't change over time (we cannot safely assume that a URL pattern is a strong enough guarantee of immutability); so there is no way to ensure automatically that your contribution matches the general rule of not changing its content with time. If we want a more agile SimRel without any risk on quality, we need to be able to check that automatically.
Presence of specified versions can be easily checked automatically, and they do ensure that if build completes (if content were not removed), what's included by your contribution is exactly the same over time. This is a reason why mentioning feature versions is highly superior to relying on p2 repository contracts or assumptions. Another reason is that the aggregator output is *what* you contribute (the artifacts) not *from where* you contribute stuff, so the mapping from SimRel model to aggregator output is more explicit and straightforward obvious if versions are set, as it doesn't rely on an effort of version resolution.
The redundancy is not the cause of any issue from SimRel perspective, the cause of irreproducibility/mpredictability is actually the lack of details (versions) in the contributed b3aggrcon files. The redundancy is mainly an issue for us people being fairly "lazy" (no disrespect here, laziness is the research of simpler approaches, and the simpler things are almost always the better) so we minimize our effort when changing a b3aggrcon file. I fully understand and support laziness, but I believe it mustn't be in the cost of quality or agility; I hope we'll soon have better tools to embrace laziness, quality and agility so I'm trying to put the necessary features in the b3 editor to make such updates simple enough to not be perceived as a redundant effort https://bugs.eclipse.org/bugs/show_bug.cgi?id=485472
Back to the top