A Windows user constantly gets ERROR: org.eclipse.equinox.p2.engine code=0 Can't download artifact... on Oomph install

[Mark-J-Lawrence]

I've set up an Oomph product and a couple of projects for our teams, which has been working very well - though there's always been a few random niggles that has come back from Windows users every now and then the last few years.

We've a new one I've been looking at the past few days from a Windows user, who had been using an Oomph installed IDE successfully the last 6 months. He works remotely, therefore is using a VPN to connect to the company network, which is required as I've created a couple of mirror sites on our company internal network, which our Oomph profiles choose to get all their plugins from.

He started getting these popups during startup:

After OKing these, Eclipse was restarted and he was greeted with:

After trying a few things out - we just said to try a fresh install and fingers crossed it would work. Sadly, it hasn't!

After clearing everything down with him, from a blank canvas - I noticed his installation is remarkably slower than running the install on my Windows 11 VM (could be internet speed, could be anti-virus?)

But his installs continually fail with ERROR: org.eclipse.equinox.p2.engine code=0 Can't download artifact ...... from any of the following repositories ....

Full log:
LOG-02.txt

We've had tried so many things that I feel like we've exhausted all options:

• He can download these plugins in the errors with no problems from his machine via a browser
• We've ensured VPN client settings are open & nothing is blocked
• Added an array of exclusions to the Windows Defender list (e.g. ~/.eclipse, ~/.p2, the parent eclipse-inst directory, the directory where IDEs are installed, windows temp)
• Played with the Active Provider settings under Network Connections - Native & Direct produce the same result.
• mirrors are set to false in the installer

Other Windows users in the company aren't having these problems, so it points the finger to either be something environmental on the machine or the network speed? (Mac users see zero issues)

Are there any hints or clues as to why this is happening? Thanks in advance!

[merks]

Installing from a zip is horrible; hopefully there is not a lot of content there. A major problem is that p2 is really stupid with a .zip file repository and if I recall correctly it will download the entire zip in order to get one artifact or even just the content/artifact metadata resource and will will do that for every single thing it tries to fetch out of that jar. 🤮 There is huge room for improvement on that front. Here's an example from my cache:

If there was literally a failure to download an artifact, that would be logged as a failure explicitly, but I don't see that. The error that an artifact not found in any repository is generally an indication that p2 is seeing inconsistent artifact metadata relative to the content metadata. I.e., it makes a plan, decides to install some IU a.b.c, that IU has an artifact a.c.b_key, and then it has to find that key in some artifact repository metadata, but it doesn't find it at all, so then it lists all the repositories where it tried to find it.

P2 keeps a cache in <home>\.p2\org.eclipse.equinox.p2.repository. Unfortunately the names are some horrible hash codes. But you should try to clear that cache. Oomph keeps a similar cache in <home>\.eclipse\org.eclipse.oomph.p2\cache, but these with names from which you can deduce originating URI. You can clear this cache too. This cache will be rebuilt whenever you try to install or update something. The cache should refresh properly, but you never know, so clearing it ensure that you are definitely not seeing stale results from a cache. If you know from which repository you are expecting to get the artifact, you can look in the Oomph cache to see what's going on with the artifacts jar/xml for that repository. Is it there? Does it have the expected content?

Good luck. It sounds nasty. 😢

[Mark-J-Lawrence]

Hi @merks, thanks for getting back to me with the pointers.

Regarding the zip - its 2MB in size and only has very few files within it, so it should be OK, and has been fine for all other users.

I didn't mention above, but we've tried an install directly after deleting ~/.eclipse and ~./.p2, so the caches should had been completely fresh. Regardless, I've a copy of the users .eclipse folder and I've been looking through the cache folder. Looking at the log, it had a problem with osgi.bundle,bcprov,1.78.1, knowing which repo that comes from, I've consulted the cache and both the "site artifacts" and "site content" XMLs have references to the package at this version. I think it would be expected content.

At least 2x Windows users have successfully installed over a VPN since raising this Q&A - yet this user still cannot. It's a bit of a mystery....is there anything else we can try?

update - the User has cleared out the caches after a failed install and retried, however its the same error afterwards.