Stream: t-compiler/wg-nll

Topic: blog posts


pnkfelix (Jun 28 2019 at 08:57, on Zulip):

Hey @WG-nll , just wanted to point you all at a blog post I put up last night that tries to enumerate all the ways that NLL is going to "break" peoples code: Breaking News: Non-Lexical Lifetimes Arrives for Everyone

lqd (Jun 28 2019 at 08:58, on Zulip):

it looks awesome Felix, great job

pnkfelix (Jun 28 2019 at 08:58, on Zulip):

Thanks; those gists you made @lqd summarizing the output from the crater runs were a huge help

pnkfelix (Jun 28 2019 at 08:59, on Zulip):

(because I strived to include examples that were reduced from actual regressions that I could point to)

lqd (Jun 28 2019 at 08:59, on Zulip):

I wish (and have vague plans) for a tool to help us categorize crater results much like you did in the spreadsheet

pnkfelix (Jun 28 2019 at 09:00, on Zulip):

Such a tool, if made general purpose, could indeed be quite useful

pnkfelix (Jun 28 2019 at 09:00, on Zulip):

it was super painful doing that by hand.

lqd (Jun 28 2019 at 09:00, on Zulip):

I think mark recently did a bunch of work in this similar direction for the recent beta regressions

lqd (Jun 28 2019 at 09:02, on Zulip):

thankfully it's not that often that we have to run through so many "regressions" :)

davidtwco (Jun 28 2019 at 09:05, on Zulip):

That's an excellent post @pnkfelix, really exciting to see migrate mode enabled for 2015.

lqd (Jun 28 2019 at 10:40, on Zulip):

(also, thank you for the classy shout-out to the WG at the end of the post :)

simulacrum (Jun 29 2019 at 00:18, on Zulip):

I think mark recently did a bunch of work in this similar direction for the recent beta regressions

My tool mostly just eliminates the "this crate isn't actually the root cause" -- i.e., when a dep fails to compile -- by grouping

simulacrum (Jun 29 2019 at 00:25, on Zulip):

@pnkfelix fwiw, I found the liveness check section a bit confusing -- it looked initially like T::Type should play some role since it's there, but it can actually be simplified to this. Great post overall though, good job!

pnkfelix (Jun 29 2019 at 05:24, on Zulip):

You’re right; that might have been something we fixed about impl Trait in MIR borrow-check. Not sure if that needs it’s own section (and a different example for the licenses check). Was too tired to properly verify when I wrote that part

pnkfelix (Jun 29 2019 at 05:24, on Zulip):

*liveness, not “licenses”

simulacrum (Jun 29 2019 at 12:55, on Zulip):

no worries

Last update: Nov 21 2019 at 14:35UTC