Crater runs for 1.48

[Mark-Simulacrum]

Note: Please do not conduct triage on these runs without discussing how to do so with a release team member first. Thanks!

[Mark-Simulacrum]

@craterbot run name=beta-1.48-1 start=1.47.0 end=beta-2020-10-09 mode=build-and-test cap-lints=warn p=10

[craterbot]

👌 Experiment beta-1.48-1 created and queued.
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[Mark-Simulacrum]

@craterbot run name=beta-1.48-rustdoc-1 start=1.47.0 end=beta-2020-10-09 mode=rustdoc cap-lints=warn p=5

[craterbot]

👌 Experiment beta-1.48-rustdoc-1 created and queued.
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🚧 Experiment beta-1.48-1 is now running

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🚧 Experiment beta-1.48-rustdoc-1 is now running

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🎉 Experiment beta-1.48-1 is completed!
📊 424 regressed and 53 fixed (126219 total)
📰 Open the full report.

⚠️ If you notice any spurious failure please add them to the blacklist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[Mark-Simulacrum]

I triaged the build failures from the crater run, but ran out of energy to open the ~245 test failures. I personally suspect many of them to be the uninitialized checking getting more stringent; I plan to finish this up later today hopefully.

Opened/tagged:

• could not fully normalize type regression: could not fully normalize type #78139 (build pass -> ICE)
• array methods unstable feature use regression: array methods unstable feature use #78140 (build pass -> error)
• use dep1::foo as dep1 is considered ambiguous use dep1::foo as dep1 is considered ambiguous #77586 (build pass -> error)
• target_feature no longer permitted in some places regression: target_feature no longer permitted in some places #78143 (build pass -> error)

There's also a bunch of failures in build scripts due to old versions of crossbeam-epoch having UB around mem::uninitialized, but that's intentional/won't fix. I believe those folks should be a cargo update away from fixing the failure.

[Mark-Simulacrum]

There's a bunch of segfaults/aborts etc -- 49 of them -- that I have not yet tried to diagnose, see https://gist.github.com/Mark-Simulacrum/d0759db08cd0528ae2762c1aecf04c37 for a list of log files. Beyond that I've triaged the rest of the test failures and those look good. I think for these it would be good to re-run crater on the subset and see how many of them are still present (which should be quick), I'll look at doing that shortly.

[Mark-Simulacrum]

@craterbot run name=beta-1.48-2-subset start=1.47.0 end=beta-2020-10-16 mode=build-and-test cap-lints=warn p=10 crates=list:https://gist.githubusercontent.com/Mark-Simulacrum/ecfa4deb6ba7151067d71269ba567f30/raw/c0375bdc243814a9af60f0ff19ada599091ce8ad/gistfile1.txt

Hopefully this works...

[craterbot]

🚨 Error: missing desired crates: {"https://gist.githubusercontent.com/Mark-Simulacrum/ecfa4deb6ba7151067d71269ba567f30/raw/c0375bdc243814a9af60f0ff19ada599091ce8ad/gistfile1.txt"}

🆘 If you have any trouble with Crater please ping @rust-lang/infra!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[Mark-Simulacrum]

Huh. Not sure what's wrong there. @pietroalbini would you be able to find some time to take a look? I don't think it's super urgent but would be good to have this cycleish.

[pietroalbini]

@craterbot run name=beta-1.48-2-subset start=1.47.0 end=beta-2020-10-16 mode=build-and-test cap-lints=warn p=10 crates=https://gist.githubusercontent.com/Mark-Simulacrum/ecfa4deb6ba7151067d71269ba567f30/raw/c0375bdc243814a9af60f0ff19ada599091ce8ad/gistfile1.txt

[craterbot]

👌 Experiment beta-1.48-2-subset created and queued.
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🚧 Experiment beta-1.48-2-subset is now running

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🎉 Experiment beta-1.48-2-subset is completed!
📊 21 regressed and 2 fixed (47 total)
📰 Open the full report.

⚠️ If you notice any spurious failure please add them to the blacklist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[Aaron1011]

The first segfault (ya2d2 in https://crater-reports.s3.amazonaws.com/beta-1.48-2-subset/beta-2020-10-16/gh/abhijat.ya2d2/log.txt) is caused by #71274 - an old version of crossbeam-epoch is in use. When mem::uninitialized triggers a panic, crossbeam-epoch ends up performing some raw pointer operations with the queue in an invalid state, leading to a segfault.

This is easier to see if the tests are run with cargo test -- --nocapture, since the initial panic message is otherwise hidden.

[craterbot]

🎉 Experiment beta-1.48-rustdoc-1 is completed!
📊 207 regressed and 1 fixed (126219 total)
📰 Open the full report.

⚠️ If you notice any spurious failure please add them to the blacklist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[RalfJung]

See this hackmd for the previous regression analysis of #71274. The vast majority of these cases has semver-compatible updates (in particular, everything caused by linked-hash-map), but I am not sure if crater has picked those up -- in the past I had trouble with crater using a broken version of a crate even when a fixed version was available.

We focused on crates though and mostly skipped GH repositories; in my prior experience, 100% of those regressions were one-off repositories of unmaintained code (things like advent of code) or the cause was in some dependency on crates.io, so there was little expected payoff from going over all of those as well. So there's nothing on ya2d2 in the hackmd.