Stream: t-compiler

Topic: log codegen regression


Steven Fackler (Nov 19 2019 at 20:00, on Zulip):

Does anyone have any intuition as to what would have caused this? https://github.com/rust-lang/log/issues/365

mati865 (Nov 19 2019 at 20:08, on Zulip):

Any clue when it started?

Steven Fackler (Nov 19 2019 at 20:08, on Zulip):

Not yet - I just noticed it a couple minutes ago. I'm going to try to get a minimal repro on godbolt and run back through their verisons

Steven Fackler (Nov 19 2019 at 20:16, on Zulip):

it looks like the existence of the fmt::Arguments argument breaks the optimization: https://rust.godbolt.org/z/ytgGyX

Steven Fackler (Nov 19 2019 at 20:16, on Zulip):

it does the right thing with that argument commented out, but the wrong thing when its present

Steven Fackler (Nov 19 2019 at 20:18, on Zulip):

oh - it's not assembling the tuple, its assembling the Arguments

Steven Fackler (Nov 19 2019 at 20:20, on Zulip):

never mind me!

Last update: Dec 12 2019 at 00:55UTC