Kamil Śliwak

@ekpyron @chriseth @axic Since this is a brainstorming issue and has been open for quite a while, I guess, if there were any more concrete ideas beyond #11436 and #11972,...

Yeah. It's not a large sample, but it was reproducible when I ran it several times (see the table) so I think it's legit. BTW, I had to resolve a...

These bugs are quite insidious and we still haven't eliminated all of them. For example I recently found #15134. Unless you're absolutely sure it's the same exact bug it would...

> Ah OK, in that case it's more likely than not that this is fixed as of the 0.8.27 (26?) release, but we'll have to check on Monday regardless. @nikola-matic...

Looks like there's already an older issue about this: #9645.

Yeah, we're aware of this particular pain point and we get a lot of complaints about it. It's one of the tasks on our roadmap for Q2 2022.

This is unfortunately blocked, pending improvements in the optimizer so we have to postpone it until we have those. That will probably only happen on top of the new type...

It's because of how the IR codegen is designed. In the legcacy codegen we'd just allocate memory and write each element already at the target location. Which sounds simple, but...

Related issue: #9592 (try/catch with low-level calls).

I think we might be better off closing this and reformulating the problem to be more focused. As is it is now, it touches upon multiple things: - Local reverts...