possible compile-time regression

[nikomatsakis]

According to perf.rust-lang.org, rollup commit 51b0b37 seems to have caused a big jump in the "Summary" numbers. However, scrolling down through the individual tests, we don't see that same jump mirrored anywhere else. Moreover, clicking on the commit in question doesn't give the usual test-by-test breakdown, just an empty page. I'm guessing this is some sort of fluke, but it'd be great to have that confirmed.

(Also, what exactly do the summary numbers represent, anyway?)

cc @Mark-Simulacrum @rust-lang/compiler @rust-lang/infra

[nikomatsakis]

triage: P-high

We need to know what happened.

[kennytm]

This is likely just rust-lang/rustc-perf#178. Nothing to see here, move along.

[nikomatsakis]

Should we close? How "likely" :)

[kennytm]

Very likely given the commit hash, but I'll let @Mark-Simulacrum confirm ☺️

[Mark-Simulacrum]

I'll investigate. Assigned myself.

[Mark-Simulacrum]

We weren't compiling crates.io in either release or debug mode because ring failed to compile, after #47298 was merged in the rollup PR crates.io began compiling, and so the summary numbers jumped up. I'm going to close this issue as nothing to be worried about. I'm working on a patch that'll allow this to be more visible in the compare view (instead of just breaking).