[phobos] freebsd64 and osx32 + druntime semaphore test == evil

[David Nadlinger]

Just to let you know, I can easily reproduce this on my OS X 10.7 box by 
simply running testWaitTimeout() function in a loop:

core.exception.AssertError at core.sync.semaphore(530): Assertion failure
----------------
5   semaphore                           0x000ff46d onAssertError + 65
6   semaphore                           0x00108886 _d_assertm + 30
7   semaphore                           0x000f7cb0 void 
core.sync.semaphore.__assert(int) + 24
8   semaphore                           0x000f84dd void 
core.sync.semaphore.testWaitTimeout() + 317
9   semaphore                           0x000f85c1 void 
core.sync.semaphore.__unittest1() + 17
10  semaphore                           0x000f7c93 void 
core.sync.semaphore.__modtest() + 11
11  semaphore                           0x000ff8e1 extern (C) bool 
core.runtime.runModuleUnitTests().int __foreachbody262(ref 
object.ModuleInfo*) + 45
12  semaphore                           0x000fac93 int 
object.ModuleInfo.opApply(scope int delegate(ref object.ModuleInfo*)) + 79
13  semaphore                           0x000ff7d2 runModuleUnitTests + 134
14  semaphore                           0x00108fda extern (C) int 
rt.dmain2.main(int, char**).void runAll() + 38
15  semaphore                           0x00108b51 extern (C) int 
rt.dmain2.main(int, char**).void tryExec(scope void delegate()) + 29
16  semaphore                           0x00108aeb main + 179
17  semaphore                           0x000f7c75 start + 53
18  ???                                 0x00000001 0x0 + 1

The assertion fails because alertedOne is not set, and about half of the 
times this happens, alertedTwo is set instead. The number of test 
iterations it takes for the first failure to happen varies from several 
tens to thousands on my machine.

David


On 8/4/11 2:41 AM, Brad Roberts wrote:
> I'm not sure why these two systems are so pissed off at the druntime
> semaphore test, but they both are, in different ways.
>
> OSX has, essentially forever, periodically crashed during the test.
>
> Recently, freebsd64 has started hanging in the test an awful lot, but not
> every time.  Eventually I notice it's been running way too long and kill
> it.
>
> I'd be happy to give access to the freebsd64 box for anyone who's willing
> to dig in.  The osx box isn't mine to give access to, but I imagine sean
> wouldn't mind adding another account should a volunteer show up to dig in
> there too.
>
> Anyone?
>
> Thanks,
> Brad
> _______________________________________________
> phobos mailing list
> phobos at puremagic.com
> http://lists.puremagic.com/mailman/listinfo/phobos