Regression report stable-2015-08-06 vs. nightly-2015-09-04


#1

Regression report stable-2015-08-06 vs. nightly-2015-09-04

  • From: stable-2015-08-06
  • To: nightly-2015-09-04

Coverage

  • 2938 crates tested: 1594 working / 1125 broken / 22 regressed / 184 fixed / 13 unknown.

Regressions

  • There are 12 root regressions
  • There are 22 regressions

Root regressions, sorted by rank:

Full report.

Analysis

This run is five days old, so some things may have changed. There’s quite a bit of breakage here, some the same as previously.

This stuff is going to hit beta next week.

  • cannot infer lifetime - nickel
  • ICE - piston-gfx_texture, glium_text, lux
    • supposedly fixed on master. I’ll make another run now
  • read_exact became unstable? - pod, immeta, rcon, byteorder, terminal_thrift
    • I don’t understand what’s going on here. Compiles on stable, but read_exact unstable on master.
  • wrong number of lifetime parameters - influent
    • no idea
  • unresolved std::hash::hash import - cuckoofilter
  • unused_mut lint - rustache
    • acceptable lint change

The read_exact thing is a mess. @bluss suggests it’s a conflict with custom implementations that already existed.


#2

This reproduces the influent regression:

trait Tr {}

struct S<'a>(&'a str);

fn f(a: S, b: Box<Tr>) -> S { a }

Works on stable, not on nightly. Using another S instead of the Box<Tr> always fails, as expected.


Regression report stable-2015-08-06 vs. nightly-2015-09-08
#3

That testcase looks like https://github.com/rust-lang/rust/issues/27248 .


#4

Thanks for the tips @jschievink @eefriedman. I’ll investigate whether we can get that fixed.