On Wed, 20 Nov 2019 at 10:54, Laurent Vivier <laur...@vivier.eu> wrote: > I understand your point, and I agree, but not allowing this will block > the merge of the hexagon target, and I don't see any fix for the > underlying problem coming soon. > > Other targets work without this change, and adding this change breaks > some user space applications (like go), whereas adding this change for > hexagon target only will improve the situation for it (with no > regression, as it doesn't work at all for the moment)
I care more that we should fix things correctly and maintain the consistency of how our architectures behave than that we are able to quickly land a target for a fairly minor architecture, to be honest. If we land hexagon with hacks and workarounds then we're potentially stuck with that behaviour. thanks -- PMM
