Upgrade the Travis CI macOS Xcode version to 8.3

[kennytm]

No description provided.

[kennytm]

@bors try

(PR queue is too crowded. Sorry for abuse.)

[bors]

⌛ Trying commit 508f9ca with merge 0cf6ddf...

[kennytm]

@bors try-

[kennytm]

@bors force clean retry

Wat, shouldn't try- cancel the build as well?!

[kennytm]

xcode9 still results in timeout: https://travis-ci.org/rust-lang/rust/builds/272110207

Next, xcode8.3.

[kennytm]

xcode8.3 looks good! https://travis-ci.org/rust-lang/rust/builds/272176590

• The x86_64 one timed out normally after 3 hours (recompiling LLVM takes most time)
• The i686 one 30-minute-no-output timed out due to sccache issue, let's disregard that.

LLVM needing 50 minutes to rebuild might be an issue though. I'm not sure if the dedicated VM for auto is faster than the VMs for PRs.

[kennytm]

r? @alexcrichton

Summary of the experiment:

• Xcode 9 beta 6 still timed out. Additionally, LLVM's feature detection wrongly assumed the host supported futimens, which is only available on the unreleased macOS 10.13, which caused a linker error when building librust_llvm.

• Xcode 8.3 seems to work fine, assuming the LLVM build time can be built faster than 50 minutes.

[alexcrichton]

@bors: r+

Thanks for the investigation here @kennytm! I imagine that once the sccache caches are populated the LLVM build time should go way down.

[bors]

📌 Commit a6cd374 has been approved by alexcrichton

[kennytm]

@bors p=8 (might reduce spurious failure)

---

Note: If any of the two check jobs timed out due to 30 minutes of inactivity, and there is no obvious cause of failure, it means the fix does not really work. Please don't retry in these cases.

[bors]

⌛ Testing commit a6cd374 with merge 85b48bd37ed5833b559748aeb6ca1e290e293447...

[bors]

💔 Test failed - status-travis

[kennytm]

It is the dist DEPLOY_ALT=1 job timed out this time 🤔

Building stage0 compiler artifacts (x86_64-apple-darwin -> x86_64-apple-darwin)
[00:13:02] warning: a `-j` argument was passed to Cargo but Cargo is also configured with an external jobserver in its environment, ignoring the `-j` parameter
...
[00:13:46]    Compiling flate2 v0.2.19
[00:16:11]    Compiling proc_macro v0.0.0 (file:///Users/travis/build/rust-lang/rust/src/libproc_macro)
[00:16:11]    Compiling rustc_const_math v0.0.0 (file:///Users/travis/build/rust-lang/rust/src/librustc_const_math)
[00:16:22]    Compiling syntax_ext v0.0.0 (file:///Users/travis/build/rust-lang/rust/src/libsyntax_ext)

No output has been received in the last 30m0s, this potentially indicates a stalled build or something wrong with the build itself.
Check the details on how to adjust your build configuration on: https://docs.travis-ci.com/user/common-build-problems/#Build-times-out-because-no-output-was-received

The build has been terminated

---

@bors retry — with the assumption that dist job failure at xcode7 is much rarer than check job failure at xcode8.2.

[bors]

⌛ Testing commit a6cd374 with merge c115f1a...

[bors]

💔 Test failed - status-travis

[kennytm]

dist x86_64-apple-darwin timed out after 3 hours, due to usually long time spent on building libstd. Everything else green-checked 🎊. (It could also be the issue is not happened in these hours.) Meanwhile Travis is investigating an incident around macOS builds, which could also affect the outcome.

Time (in minutes) spent on `dist x86_64-apple-darwin`

	58632.8	58618.8
make prepare	3.8	2.2
stage0-std	26.3	1.2
llvm	6.3	5.7
stage0-rustc	34.6	16.8
stage1-std	3.3	3.0
stage1-rustc	39.0	21.2
stage2-std (→ x86_64-apple-ios)	25.0	1.3
stage2-cargo	timed out	9.2

 

Currently reducing priority back to normal to check the behavior of other PRs first. If two other PRs can pass normally than we don't need PR.

@bors retry p=0

[kennytm]

@bors p=8

[bors]

⌛ Testing commit a6cd374 with merge db86d16b833aba9cca7b63bcf63a174bf1411fb9...

[bors]

💔 Test failed - status-travis

[kennytm]

So that shows the Xcode version is likely irrelevant to the actual problem.

Since Travis is already investigating the cause, hopefully it will be fixed by upstream.

@bors p=0 rollup

[kennytm]

The incident should be resolved. Since Xcode 9 is expected to be released in a few weeks, as a regular PR it is better to just upgrade to Xcode 9-stable directly after Travis supported it. Closing.