geronimo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Joe Bohn <joe.b...@earthlink.net>
Subject windows build hell
Date Wed, 30 Aug 2006 15:36:44 GMT

We've been struggling with build problems on windows for some time now 
and things just keep on getting worse.

I'm composing this note to summarize the problems that I'm aware of with 
work-arounds and possible solutions.   I'm also fishing to see if 
anybody else has some ideas on how to resolve these issues.

1)  Windows pathlength problem.
Problem:
There are miscellaneous failures that can typically be tracked down to 
the windows pathlength problem.  The nature of our repository structure 
and deployment make this a big problem.   It also seems like we're 
continuing to add more intermediate elements in path names as we try to 
get things more organized and conform to Maven 2 conventions.

Work-around:
The work-around is to keep the windows root path as small as possible. 
I now typically build from a root path of c:\g to avoid these problems.

Possible Solution:
For a longer term solution I'm planning to work on a new repository 
implementation in 1.2 that that isn't as redundant or verbose.



2)  JSP compilation errors
Problem:
Embedded error: Unable to compile class for JSP
Strange error message about JAVA_HOME, etc...

Possible Solution/Work-around:
Update the pom.xml in the root directory to use version 1.4.5-SNAPSHOT 
(from 1.4.4) for the jspc-maven-plugin.  Not sure if Jeff Genender is 
planning to make 1.4.5 an official release for this.  We're not sure why 
it gets us around the problem so it may be a red herring.



3)  Openejb2 test failures.
Problem:
Caused by: java.lang.NoSuchMethodException: 
org.openejb.deployment.DeploymentTestSuite.getName()

Work-around:
After bootstrap failure cd to <root>\target\external\openejb2
"mvn -Dmaven.test.skip=true"
cd back to <root>
"mvn clean install"

Possible solution:
Dain suggested adding the "getName" method to the test.  However, when I 
attempted this I hit other errors.
http://marc.theaimsgroup.com/?l=geronimo-dev&m=115680051431478&w=2
I think it would be helpful if we could disable the openejb tests until 
this problem is resolved.



4)  Blue screen of death (bsod)
Problem:
This has been reported by multiple users on various machines.  When 
running an M1 or M2 build the user encounters a bsod due to a memory 
failure.
PAGE_FAULT_IN_NONPAGED_AREA
  ***  STOP:  0x00000050 (0xBADDB148, 0x00000000, 0x8056C77B, 0x00000000)
Dump of physical memory

Work-around/Possible Solution:
Haven't found one yet.  I've tried updated drivers, replaced hardware, 
tried various heap size settings, etc....  At times this can be fairly 
frequent (every 3rd build attempt or so).  I'm collecting bootstrap.log 
files from folks when this happens during a bootstrap to see if there is 
a common thread.  So far, with the bootstrap logs, it always seems to 
happen at about the same place:
Running tests after building a module (usually tomcat.ApplicationTest or 
TomcatModuleBuilderTest) and the final lines in the log are always the 
Creation of an MBeanServer like this:
  [exec] Running org.apache.geronimo.tomcat.ApplicationTest
  [exec] Created MBeanServer with ID: 5dcec6:10d5a184aed:-8000:jbohn2:1

Please respond to this note if you are seeing the bsod failures on 
windows.  At first I thought this was just me and was hardware related. 
  However, the more I talk to folks on windows the more I hear of other 
folks encountering this same problem.  I've updated all drivers, 
replaced my entire system, and several other folks have reported seeing 
this on completely different systems.   I think that pretty much rules 
out a hardware problem.

ideas welcome!

Joe









Mime
View raw message