On Fri, 23 Apr 2021, at 04:22, Håvard Ottestad wrote:
Hi,
I’ve been working on a major performance issue with the rsx:targetShape implementation that I’m getting close to merging.
I want to release this as soon as possible with 3.6.3.
No prob from my point of view. I'm currently mostly focused on stuff for 3.7 or beyond.
I see that the OOM issue for the overflow model is scheduled for 3.6.3. Would it be ok if I tried adjusting the various limits in the overflow model to see if we can get a quick fix that at least stops users from experiencing OOM errors?
By all means have a go, though as said on the issue it is unlikely that it is something we can fix this by only adjusting the threshold. I'm happy to merge with a lower threshold as a quick workaround, to at least make the chance of an OOM smaller, but to really fix the problem we should probably consider a substantial refactor of the MemoryOverflow model.
Is there anything else we want to include in 3.6.3?
There's another open issue on 3.6.3: GH-2749. We have a third party contributor who's offered to look into it but I just realized they had a question a month ago, and I completely overlooked it. I've responded to them now, I hope we can still get them to put up a PR, but it's not a blocker for 3.6.3.
Eyeballs on my PR are also very welcome :)
I have had a look but it's a bit too specialized for me to follow what's going on. There's one question I have (related to the number of test case files that has changed), but apart from that it looks ok to me, as far as I can judge.
Jeen
_______________________________________________rdf4j-dev mailing listrdf4j-dev@xxxxxxxxxxxTo unsubscribe from this list, visit https://www.eclipse.org/mailman/listinfo/rdf4j-dev