Stream: t-compiler

Topic: Removing `mir_const_qualif` query

ecstatic-morse (Oct 27 2019 at 01:53, on Zulip):

After re-reading the rustc guide, I think this is an external vs local provider issue

Matthew Jasper (Oct 27 2019 at 09:36, on Zulip):

You could have mir_const_qualif get the qualifs from mir_validated and then you would only have to encode that.

ecstatic-morse (Oct 27 2019 at 17:21, on Zulip):

Ah, yes, I'll give this a try. Thank you!

eddyb (Oct 30 2019 at 12:50, on Zulip):

that's... one way, I guess? in general, I wish we had first-class support for "one provider, multiple queries" (cc @nikomatsakis)

nikomatsakis (Oct 31 2019 at 12:54, on Zulip):

Yes -- we had some thoughts about this related to how we handle MIR lowering as well that might actually be relevant

nikomatsakis (Oct 31 2019 at 12:54, on Zulip):

I should try to write them down

ecstatic-morse (Oct 31 2019 at 17:47, on Zulip):

BTW, removing mir_const_qualif was an attempt to fix some bugs around promotion for array initializers in consts while keeping the existing const-checking/promotion pass. Once the combined pass is removed, two separate queries will once again be the best strategy.

Last update: May 24 2020 at 23:40UTC