Sam Hokin wrote:
Caldarale, Charles R wrote:
Once you have your test Tomcat set up on the problematic server, try
adding this line to the conf/logging.properties file:
org.apache.jasper.level = ALL
That will display a *lot* of information about what's going on during
JSP compilation. One thing in particular printed out during Tomcat
initialization is the compiler classpath for JSPs in each webapp,
which might show why the lookups are being done at the root of the
file system.
OK, done. I started up my test instance, ran a single request on my
test JSP, and shut it down. The resulting
catalina.out is here for your review:
http://ims.net/media/jasper-level-all.txt
Sure enough, / appears in the classpath for the ROOT webapp, right here:
[stuff]:/usr/local/lib/jtds-1.2.2.jar:/:/usr/local/tomcat-2/bin/bootstrap.jar:[other
stuff]
/ is appended after a jar that I have symlinked under $TOMCAT_HOME/lib
and before bootstrap.jar.
and, this is also maybe where these lstat calls were coming from (lstat
varies from stat in that it queries the symlink itself, rather than the
underlying real file).
So, SINCE my company is ims.net and my Java packages start with net.ims;
and SINCE I had autofs running which
automounted /net; and SINCE / is in the classpath, JDT searched under
/net to find core classes, which incurred a very
long delay, presumably due to automount. If autofs is off and /net is
simply an empty directory, the delay is not
incurred. My problem is cured by simply turning off autofs.
It seems to me that it's a bug that / is being appended to the
classpath, but one that didn't result in pain until the
particular circumstance above came into play, clearly a VERY specific
scenario.
THANKS, EVERYONE! I think we've solved this riddle! I owe you all beers,
or your beverage of choice!
My preferred one is Westvleteren, arguably the best beer in the world.
See http://en.wikipedia.org/wiki/Westvleteren_Brewery
Cheap, but hard to get.
You can have it delivered at the ApacheCON Europe 2009, and I will take
care of the distribution to deserving contributors. Maybe we can even
get some to Minnesota, along with some Belgian dark chocolate.
I am very glad that this turned out after all as a (fruitful)
intellectual pursuit, rather than a boring hardware issue.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org