I'm getting random "Home directory does not exist" errors when building the core-model-tests. The thing is, its random. 4-5 times it fails, 1-5 times it succeeds. So, I put a 10 millisecond delay after any mkdir() call in TestModelControllerService and it seemed to solve the problem. Dont' know if its my machine, Windblows or what. Maybe just a race condition between JVM and the OS's file controller.

Not anymore.

Anything that required mkdir() calls was failing with java.io directory not found errors. Again, the 10 millisecond delay fixed the problem.

On 6/3/2013 12:09 PM, Brian Stansberry wrote: > Do you have test output from the failures? > > On 6/3/13 9:08 AM, Bill Burke wrote: >> I'm getting random "Home directory does not exist" errors when building >> the core-model-tests. The thing is, its random. 4-5 times it fails, >> 1-5 times it succeeds. >> >> So, I put a 10 millisecond delay after any mkdir() call in >> TestModelControllerService and it seemed to solve the problem. Dont' >> know if its my machine, Windblows or what. Maybe just a race condition >> between JVM and the OS's file controller. >> > >

This looks really fishy. Given that I use windows as primary OS and we do test extensivly wildfly on windows build agents. Problems like this never showed up before. I would be almost certain that the problem has to be related to your enviroment. For example I run Windows 8 64bit with SSD and also tested on Win7 with SSD and there was no problem at all. on build server we test on Windows Server 2008r2 (same kernel as Windows7) and on Windows Server 2012(same kernel as windows 8) And there was never anything similar to be found. I would rule it out to your hardware/software setup, Maybe some bad driver, or faulty raid controller, btw do you use SW or HW raid? -- tomaz On Mon, Jun 3, 2013 at 7:35 PM, Brian Stansberry <brian.stansberry(a)redhat.com <mailto:brian.stansberry@redhat.com>> wrote: On 6/3/13 11:17 AM, Bill Burke wrote: > Not anymore. Ok, no problem. > Anything that required mkdir() calls was failing with > java.io <http://java.io> directory not found errors. > > Again, the 10 millisecond delay fixed the problem. > Sure, but it's nice to understand the problem. But analyzing the code I see that the "Home directory does not exist" message can only come from the same thread that did the mkdir(), which tells me what I wanted to know. > On 6/3/2013 12:09 PM, Brian Stansberry wrote: >> Do you have test output from the failures? >> >> On 6/3/13 9:08 AM, Bill Burke wrote: >>> I'm getting random "Home directory does not exist" errors when building >>> the core-model-tests. The thing is, its random. 4-5 times it fails, >>> 1-5 times it succeeds. >>> >>> So, I put a 10 millisecond delay after any mkdir() call in >>> TestModelControllerService and it seemed to solve the problem. Dont' >>> know if its my machine, Windblows or what. Maybe just a race condition >>> between JVM and the OS's file controller. >>> >> >> > -- Brian Stansberry Principal Software Engineer JBoss by Red Hat _______________________________________________ wildfly-dev mailing list wildfly-dev(a)lists.jboss.org <mailto:wildfly-dev@lists.jboss.org> https://lists.jboss.org/mailman/listinfo/wildfly-dev _______________________________________________ wildfly-dev mailing list wildfly-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/wildfly-dev

I hope this would not be the final solution. Otherwise next time we start testing on some poor machine in the cloud we may start seeing transient failures. Bill Burke wrote, On 06/03/2013 05:08 PM (EEST): > I'm getting random "Home directory does not exist" errors when building > the core-model-tests. The thing is, its random. 4-5 times it fails, > 1-5 times it succeeds. > > So, I put a 10 millisecond delay after any mkdir() call in > TestModelControllerService and it seemed to solve the problem. Dont' > know if its my machine, Windblows or what. Maybe just a race condition > between JVM and the OS's file controller. > _______________________________________________ wildfly-dev mailing list wildfly-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/wildfly-dev

We often need to test in non-ideal environments especially older machines with proprietary software and cloud. Transient failures tend to lose a lot of time especially when they happen to fail consistently in particular environment. Not only QE time but also dev time. Also test suite becomes hard to run by inexperienced with the particular test suite people, e.g. self-certification and certification services become a hard goal. That's why I think that, as stupid as it sounds, such a loop is better than dealing with this when least expected. Does this class throw in your environment? If it works then perhaps the problem is an actual bug in test controller (it works for me on fedora 18). And if it fails then it must be a bug in the JVM.. at least the way I understand mkdir() javadoc. package com.example.test; import java.io.File; public class MkdirTest { public static void main(String[] args) {; for ( int i=0; i<100; i++) { File newDir = new File(System.getProperty("java.io.tmpdir"), "mytestdir-" + Long.toHexString(System.nanoTime())); newDir.mkdir(); newDir.list(); newDir.delete(); } System.out.println("Test run successfully"); } } Bill Burke wrote, On 06/03/2013 07:24 PM (EEST): > What would be the final solution? Loop until the directory exists? :) > I don't think so... Gotta be a delay. > > On 6/3/2013 12:15 PM, Aleksandar Kostadinov wrote: >> I hope this would not be the final solution. Otherwise next time we >> start testing on some poor machine in the cloud we may start seeing >> transient failures. >> >> Bill Burke wrote, On 06/03/2013 05:08 PM (EEST): >>> I'm getting random "Home directory does not exist" errors when building >>> the core-model-tests. The thing is, its random. 4-5 times it fails, >>> 1-5 times it succeeds. >>> >>> So, I put a 10 millisecond delay after any mkdir() call in >>> TestModelControllerService and it seemed to solve the problem. Dont' >>> know if its my machine, Windblows or what. Maybe just a race condition >>> between JVM and the OS's file controller. >>> >> _______________________________________________ >> wildfly-dev mailing list >> wildfly-dev(a)lists.jboss.org >> https://lists.jboss.org/mailman/listinfo/wildfly-dev >> > _______________________________________________ wildfly-dev mailing list wildfly-dev(a)lists.jboss.org https://lists.jboss.org/mailman/listinfo/wildfly-dev