Logging broken ?

Wed Apr 6 11:32:00 GMT 2005

Ranjit Mathew writes:
 > On Apr 6, 2005 4:34 PM, Andrew Haley <aph@redhat.com> wrote:
 > > Andreas Grunewald writes:
 > > > Thanks for the information that I had to put this into my path.
 > > > But it didn't really help much, this is my output.
 > > >
 > > > NNow starting to LOG
 > > > This is tLog java.util.logging.Logger@d68bb8
 > > > Exception in thread "main" java.lang.NullPointerException
 > > > at __ZN5Hello2goEv
 > > > (E:2004円-development2005円-extraextra\src\Hello.java:23)
 > > > at __ZN5Hello4mainEP6JArrayIPN4java4lang6StringEE
 > > > (E:2004円-development2005円-extraextra\src\Hello.java:13)
 > > > at _main (C:\DOKUME~1\GRUNEW~1\LOKALE~1\Temp\ccmocaaa.i:11)
 > > 
 > > But this is *tremendously* useful information! It tells us exactly
 > > where the NullPointerException happens.
 > 
 > Doesn't the old stack trace infrastructure prune out addresses for
 > which it cannot get proper information from addr2line? If so, the
 > stack trace could actually be misleading.
Yeah, thanks. I didn't want that change to go in, as you can see from
my comments at the time. I lost. :-(
 > The unadorned stack trace had many more entries if you notice. You
 > had to set the "gnu.gcj.runtime.NameFinder.remove_unknown" system
 > property to "false" to get a complete trace.
 > 
 > Andreas, to do this, you need to use
 > something like:
 > 
 > gcj -Dgnu.gcj.runtime.NameFinder.remove_unknown=true \
 > --main=Hello Hello.java
 > 
 > BTW, for some reason c++filt is not displaying demangled names
 > properly here. This stack trace can be filtered through "c++filt
 > -s java --format java --strip-underscores" to get a better trace.
 > 
 > > There are only two things in there that might raise a
 > > NullPointerException: Level.SEVERE, and the call to tLog.log.
 > 
 > Since tLog was (successfully) referred to a couple of lines before
 > line #23 and SEVERE is a final constant, this makes me suspect this
 > stack trace.
Could be. The only real way to find this is with gdb.
Andrew.