Name: rlT66838 Date: 07/11/97
All versions of the JDK 1.1.x (x = 1,2,3)
that I have tried on NT and Solaris SPARC
*all* have bugs in -prof. Most recently,
jdk 1.1.3 fixed the crash in java_g.
It *seems* that small programs now
profile OK in java_g. In JDK 1.1.3,
I have not seen anything wrong in the
gc part of the output. It still shows
troubles in the CPU part.
I have a profile that has 636 entries (in the cpu profile
section) and it seems OK. I have another that has
1234 entries. In this one java_g profiles
properly, but runs differently than java (it
fails inside of the jdbc-odbc bridge with
a bad connection that does not occur using java).
I have another that has 3834 entries, and things
are *missing*. It would seem that some entries just
disappear. It is easy to tell that
things are broken: there are not enough
entries for "main". The bad trace is about 1/2
MB in size.
This problem never occured in JDK 1.02.
It is new in JDK 1.1.x.
I observe it both on NT and Solaris SPARC.
The Solaris I have is not typical: it has Solaris
2.6 Beta and Java Beta on it.
----
1. Steps: Run any large Java application
under java_g and -prof. Look at the output
and see if there are enough entries for "main".
2. Sorry. We are in the process of doing
a cross licence with Sun, and this should
allow Sun to run our software (at our site).
We are just down the road.
3. N/A
4. N/A
5. N/A
======================================================================
- duplicates
-
JDK-4069612 (1.1) 'java -prof' output is incomplete (Win32 only)
-
- Closed
-