Regression report beta-2015-04-03 vs. beta-2015-04-17

[brson]

This time we’re comparing the first beta of the cycle, 1.0.0-beta, to the most recent, 1.0.0-beta.2, so this covers two weeks of changes since the beta began.

This time there are 6 reported regressions, compared to 2 last week. Notably, leveldb is no longer reported as regressed; I think this is because they’ve published a new revision, and this test is only for the most recent revisions.

The reasons for the regressions are assorted:

• Conflicting implementations of core::convert::From - mm_image.
• Thread::scoped made unstable - redlock, redlock-rs.
• More restrictive fat pointer conversion - rust_events.
• Some scalar conversion appears to no longer work - expecttest.
• us suffix - rustyham.

Coverage

• 1060 crates tested: 370 working / 676 broken / 6 regressed / 8 fixed.

Regressions

• There are 6 root regressions
• There are 6 regressions

Root regressions, sorted by popularity:

• mm_image-0.1.3 (before) (after)
• expectest-0.1.1 (before) (after)
• redlock-0.3.0 (before) (after)
• redlock-rs-0.2.1 (before) (after)
• rust_events-0.8.1 (before) (after)
• rustyham-0.0.1 (before) (after)

Full Report.

[daboross]

Just noting here, not sure if it matters:

My (binary) crate was also broken, but by the addition of a type parameter to JoinHandle. Any code which returned a JoinHandle was broken. Easy fix changing JoinHandle to JoinHandle<()>, but it is a noticeable breaking change.

[bluss]

This was duly noted in the breaking changes log (which is just a filtered git log), but with the then mistaken remark that no code was expected to break. So, @daboross, the contributors usually have a good grasp of what is a breaking change, but I’m sure the feedback helps to gauge what changes matter for real code.

[jgallagher]

I believe some of the failures are caused by non-Rust problems; e.g., libsqlite3-sys fails because it uses pkg-config to find libsqlite3, which fails.

Is this something that could / should be fixed, or do you think it’s a small enough effect that it’s not worth worrying about?

[brson]

@jgallagher It definitely should be fixed! My short term plan is to set up the docker image crater uses to contain the native deps for a bunch of the most popular stuff.

[system]

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.