Intermittent repl test failures

[staticfloat]

I'm getting intermittent test failures on OSX for the repl test:

...
        From worker 4:       * floatapprox
        From worker 4:       * readdlm
        From worker 5:       * regex
        From worker 3:       * float16
        From worker 5:       * combinatorics
        From worker 3:       * sysinfo
        From worker 5:       * rounding
        From worker 5:       * ranges
        From worker 3:       * mod2pi
        From worker 3:       * euler
        From worker 3:       * show
        From worker 4:       * lineedit
        From worker 3:       * replcompletions
        From worker 4:       * repl
        From worker 3:       * test
        From worker 3:       * examples
Worker 4 terminated.
ERROR: ProcessExitedException()
 in wait at ./task.jl:279
 in wait at ./task.jl:189
 in wait_full at ./multi.jl:594
 in remotecall_fetch at multi.jl:696
 in remotecall_fetch at multi.jl:701
 in anonymous at task.jl:1348
while loading /Users/sabae/tmp/julia-packaging/osx10.7+/julia-master/test/runtests.jl, in expression starting on line 46

@Keno any ideas of what I could do to debug this? The backtrace isn't particularly helpful. Note that this is mostly happening when I'm running tests inside Homebrew or the nightly build process, I haven't been able to trigger it reliably yet.

[staticfloat]

Hmmm. Looks like it's not just OSX either.
https://travis-ci.org/JuliaLang/julia/jobs/30194012

[tkelman]

Related to 3e2a2fa maybe?

[jiahao]

Just ran into this too on Travis: https://travis-ci.org/JuliaLang/julia/jobs/30245693#L1517

[vtjnash]

Related to 3e2a2fa maybe?

no, that one would result in a hung task

[JeffBezanson]

This is failing quite frequently on travis. It's hard to get a passing build. We really need to fix this.

[Fedster]

Hi,

failing again on the latest build:

https://gist.github.com/pao/65690e5bfbbc8c88eb39

I'm on OSX, 10.9.4.

Best

F

[pao]

@Fedster Thanks for the report. Since this is the same failure as in the original report, and I don't believe it adds any new information, I have moved the contents into a gist (available at https://gist.github.com/pao/65690e5bfbbc8c88eb39 and linked in place above) which keeps this issue's discussion concise.

[Fedster]

@pao thanks for doing that

[staticfloat]

I've got a bad feeling about this. </starwars>

Looks like it's not necessarily related to the repl test at all.

[Fedster]

Just did git fetch && git merge origin

...
From git://github.com/JuliaLang/julia
9299b25..e874beb master -> origin/master
Updating 9299b25..e874beb
Fast-forward
doc/manual/control-flow.rst | 4 ++--
doc/stdlib/linalg.rst | 14 +++++++-------
2 files changed, 9 insertions(+), 9 deletions(-)

and testall passes

[staticfloat]

Try running it a couple times. I find it's rather intermittent.
-E

On Tue, Jul 22, 2014 at 11:29 AM, Fedster notifications@github.com wrote:

Just did git fetch && git merge origin

...
From git://github.com/JuliaLang/julia
9299b25 https://github.com/JuliaLang/julia/commit/9299b25..e874beb
e874beb master ->
origin/master
Updating 9299b25 9299b25..
e874beb e874beb
Fast-forward
doc/manual/control-flow.rst | 4 ++--
doc/stdlib/linalg.rst | 14 +++++++-------
2 files changed, 9 insertions(+), 9 deletions(-)

and testall passes

—
Reply to this email directly or view it on GitHub
#7640 (comment).

[Fedster]

I just tried twice more and testall passed all the times (i.e. passed 3 times in a row).

[vtjnash]

https://gist.github.com/vtjnash/837ec5dc9e06a68a99a9
https://gist.github.com/vtjnash/14235f4ff896cdcdfd2f

i find that failures in the repl test don't give an error log however

[vtjnash]

this is the closest I got to what might be the backtrace:
https://gist.github.com/vtjnash/d02a9a275d0dbbad89ee

[vtjnash]

@tkelman is pretty sure this fixed now. i agree that not having functions lying around containing invalid pointers is probably a good thing for avoiding segfaults.

[staticfloat]

I'm still getting failures on Linux which look suspiciously similar (This log is a Linux build of 4448e7c). I have a different error on OSX now, but I'm not certain it's related.

[tkelman]

Spawn segfault, 32 bit? Seems like a different problem to me. At 4448e7c on 64 bit Ubuntu 14.04 I got 10 successful make testalls in a row. Reverting to 0195310 fails the repl test the first time (still collecting more runs).

[tkelman]

On the same machine at 0195310, ten runs of make testall gave 6 repl test failures. Looks much better now to me, but if anyone still sees repl failures on current master, speak up.

[staticfloat]

Having slept on it, I believe @tkelman is right. Thanks, @vtjnash for hunting that down!