Re: [cross-project-issues-dev] [cbi-dev] Building SimRel with Tycho for

[Jonah Graham <jonah@xxxxxxxxxxxxxxxx>]

Oops I answered Ed's email without reading the below which had already answered Ed's questions.

I have added more thoughts:

On Fri, 27 Nov 2020 at 04:21, Mickael Istria <mistria@xxxxxxxxxx> wrote:

On Fri, Nov 27, 2020 at 9:54 AM Ed Merks <ed.merks@xxxxxxxxx> wrote:

Because of that separation, goodness knows if what think I contribute will really come from my repo and not some other repo. [...] E.g., when I contribute Oomph, I do not contribute the http client bundles (as an example).  I do it that way so that I don't contribute conflicting or older versions of those things.

If I'm not mistaken, current aggregator has never guaranteed anything about provenance and does some dependency resolution against *all* repos already. I'll work on an example to prove me wrong or false and report the results and example. That will either strengthen or invalidate this argument.

You are not mistaken. The output of recent builds of simrel can show this quite clearly as it shows which repo items are being mirrored from. 

 

But what stops the Tycho resolution mechanism from grabbing potentially older the ones from the Oomph repository anyway?  Nothing I think...

Nothing indeed, unless some other requirements enforces a specific version of the artifact in its dependency chain. But as mentioned, I believe it's the same for current aggregator already. More details to come next week.

I agree.

 

I also wonder too if the two different implementations (resolution algorithms) use to build the repositories really do actually produce the same repositories?

 They're both relying on p2 director so there shouldn't be a too big difference; but we'll diff the output.

The place things *may* be different is the order of processing. I don't know if it is by design or happenstance, but b3 mirrors from the top of the .aggr file down. So if the same versioned jar is available from multiple places, the first p2 site in the .aggr file will be used.

 

The Planning Council ought to speak out and the simrel participant opinions (if any one has one!), ought to be considered too.

I agree. However, I think there are still technical questions that need an answer before we require Planning Council to voice an opinion.

 OK.