Rust 1.15 release status


#1

Hey.

The 1.15 release happens on Feb 2. Things are looking pretty good right now, but there are a number of regressions. This cycle I would like to have the beta in good shape 2 weeks ahead of time so we are not under so much pressure at the last minute. That is this Friday. If there are things you want in the beta, please get them completed.

Here are the release notes. It’s going to be a meaty release.

Here are the outstanding beta regressions. There’s 7 as of now and some have not had much attention.

There are two stable regressions that I am hoping to fix myself soon:

There are only two outstanding backports. That’s amazing. Thanks @nikomatsakis and @alexcrichton for getting those done.

Finally, let’s stay on top of nightly regressions so they don’t hit beta. There are 6 right now.

Don’t forget release triage is every other Thursday, at 9 AM PST, and is happening this Thursday.


#2

Oh, actually if I look at the calendar, 2 weeks is next Monday, the 16th, which is two weeks before we must begin the build.


#3

#38960 needs a backport, which I don’t see yet. The patch applies trivially to beta – should I open a PR?


#4

It’s beta-accepted, so yes.


#5

@cuviper Yes please!


#6

See #39008, and @alexcrichton already merged it. :slight_smile:


#7

Hey again.

The 1.15 release is looking really solid. Thank you everybody for staying on top of it, fixing and reviewing regressions, and making backports.

There are only two outstanding backports, which I will backport today:

Once these land there will be no known beta regressions. That’s great. There is also only 1 new stable regression this cycle, a rustup-related bug I still aim to investigate.

One area where we’ve struggled the last few releases is with platform breakage, so if you are using tier 2 platforms you might check that beta is still working.

And there are lots of good features this release, so it’s looking great!


#8

I’ve got 1.15.0-beta.3 building on Fedora rawhide – main build for most arches and s390x is separate. It takes a while for some to complete, especially armv7hl, but the faster ones ought to be done by now. Instead I’m seeing:

test [run-pass] run-pass/env-funky-keys.rs has been running for over 60 seconds

That’s the current state on x86_64, ppc64, and s390x at least. Looking at this test, it could get into an infinite exec loop if the arg count didn’t work for some reason. Rawhide has a prerelease glibc, so it could be something weird there too. I’ll try to reproduce this locally and file a bug…

edit: I’m also aware that beta.3 is not current with the beta branch, but nothing new there looks relevant.

edit2: I confirmed this in a local rawhide root, and copying that exact env-funky-keys binary to Fedora 25 runs fine. I’m calling this one a glibc issue for the moment… rhbz 1414589


#9

Thanks @cuviper! I imagine that test failure isn’t a blocker to actually releasing Rust 1.15 for fedora. Is that right?


#10

I don’t block the rpmbuild on test failures, but a test hang will eventually make the whole build timeout. Maybe I should mask that particular test for now, just to confirm a successful scratch build before we hit the release. Anyway, I should be able to poke the right people to get glibc fixed soon.


#11

For BSD platforms, beta has a problem when rustbuild is used. rustbuild makes assomption that “make” is GNU-make, but under BSD it isn’t the case.

It is corrected in nightly by:

The first PR contains the fix for make, and corrects also others build problems specific to OpenBSD. The second PR polish some cases for make, when makefiles are used (the first PR didn’t update that).


#12

Thank you @semarie! I will attempt to backport.