tcp-stress-test.rs fails in master

[the-kenny]

I'm packaging unstable (rev d9a911d) for NixOS. Right now I can't continue here as one test fails. I'm not sure if this is related to our chrooted build environment or if it's an upstream issue.

Noteworthy is also that beta and stable are working fine (with the same build process).

Error messages:

failures:

---- [run-pass] run-pass/tcp-stress.rs stdout ----

executing x86_64-unknown-linux-gnu/stage2/bin/rustc /tmp/nix-build-rustc-master-1.12.0-gd9a911d.drv-0/rust-d9a911d/src/test/run-pass/tcp-stress.rs -L x86_64-unknown-linux-gnu/test/run-pass/ --target=x86_64-unknown-linux-gnu -L x86_64-unknown-linux-gnu/test/run-pass/tcp-stress.stage2-x86_64-unknown-linux-gnu.run-pass.libaux -C prefer-dynamic -o x86_64-unknown-linux-gnu/test/run-pass/tcp-stress.stage2-x86_64-unknown-linux-gnu --cfg rtopt -C rpath -O -L x86_64-unknown-linux-gnu/rt
------stdout------------------------------

------stderr------------------------------

------------------------------------------
executing x86_64-unknown-linux-gnu/test/run-pass/tcp-stress.stage2-x86_64-unknown-linux-gnu
------stdout------------------------------

------stderr------------------------------
thread 'main' panicked at 'assertion failed: `(left == right)` (left: `484`, right: `1000`)', /tmp/nix-build-rustc-master-1.12.0-gd9a911d.drv-0/rust-d9a911d/src/test/run-pass/tcp-stress.rs:69
note: Run with `RUST_BACKTRACE=1` for a backtrace.

------------------------------------------

error: test run failed!
status: exit code: 101
command: x86_64-unknown-linux-gnu/test/run-pass/tcp-stress.stage2-x86_64-unknown-linux-gnu
stdout:
------------------------------------------

------------------------------------------
stderr:
------------------------------------------
thread 'main' panicked at 'assertion failed: `(left == right)` (left: `484`, right: `1000`)', /tmp/nix-build-rustc-master-1.12.0-gd9a911d.drv-0/rust-d9a911d/src/test/run-pass/tcp-stress.rs:69
note: Run with `RUST_BACKTRACE=1` for a backtrace.

------------------------------------------

thread '[run-pass] run-pass/tcp-stress.rs' panicked at 'explicit panic', /tmp/nix-build-rustc-master-1.12.0-gd9a911d.drv-0/rust-d9a911d/src/tools/compiletest/src/runtest.rs:2353
note: Run with `RUST_BACKTRACE=1` for a backtrace.


failures:
    [run-pass] run-pass/tcp-stress.rs

test result: FAILED. 2470 passed; 1 failed; 3 ignored; 0 measured

thread 'main' panicked at 'Some tests failed', /tmp/nix-build-rustc-master-1.12.0-gd9a911d.drv-0/rust-d9a911d/src/tools/compiletest/src/main.rs:293
make: *** [/tmp/nix-build-rustc-master-1.12.0-gd9a911d.drv-0/rust-d9a911d/mk/tests.mk:771: tmp/check-stage2-T-x86_64-unknown-linux-gnu-H-x86_64-unknown-linux-gnu-rpass.ok] Error 101
builder for ‘/nix/store/13ij8ra9xgqqk8xpdgv9c0hrmk189jc2-rustc-master-1.12.0-gd9a911d.drv’ failed with exit code 2
error: build of ‘/nix/store/13ij8ra9xgqqk8xpdgv9c0hrmk189jc2-rustc-master-1.12.0-gd9a911d.drv’ failed

I forgot to pipe the build output to a file, but I'm running it again now and will include a link to the full output as soon as possible.

[TimNN]

It looks like that test is already ignored on lot's of platforms:

// ignore-android needs extra network permissions
// ignore-bitrig system ulimit (Too many open files)
// ignore-netbsd system ulimit (Too many open files)
// ignore-openbsd system ulimit (Too many open files)
// ignore-emscripten no threads or sockets support

And far as I can tell it fails because it can only spawn 484 out of 1000 threads.

[TimNN]

I believe this issue manifest itself since #33640, when the failing assert was introduced.

[the-kenny]

Is this something to be handled upstream? I don't think we can easily
change the ulimit for a specific build.

Should I just patch out the failing test on our side?

Am 29.07.2016 12:53 schrieb "Tim Neumann" notifications@github.com:

I believe this issue manifest itself since #33640
#33640, when the failing assert
was introduced.

—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
#35107 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAB5v77e8BfXW7_C66vQZoKV_Joq2vSKks5qadukgaJpZM4JYAMn
.

[TimNN]

Mh, I'm not that familiar with the topic, better wait for someone like @alexcrichton to chime in, who know better, whats going on.

(Just on a side note, the // ignore-* lines were present before #33640, the PR which introduced the failing assertion).

[alexcrichton]

It should be fine to just turn down the size of this tests, it was originally meant to stress libuv and turn up bugs, but that's long gone now!

[the-kenny]

@alexcrichton Thanks! Lowering the thread count to 256 fixed the build for us. Shall we close this issue?

[alexcrichton]

Sure, sounds good to me!