Name: skT88420 Date: 12/03/99


java version "1.2.2"
Classic VM (build JDK-1.2.2-G, native threads, symcjit)

If a profiler requests a JVMPI MONITOR_DUMP or HEAP_DUMP with more
than 1024 threads, the VM panics with the message:

*** panic: too many threads

  From source inspection, this is because a fixed size buffer of size
JVMPI_DUMP_MAX_TRACES (1024) is used.

Ideally, a fixed sized buffer should not be used. Even if it is, the
VM should not panic, but indicate it could not give all the traces
(for example by giving a NULL traces pointer in the event), and this
included in the JVMPI spec.

I know 1024 threads is a lot of threads, but unfortunately a profiler
vendor does not have control over what it's users try to profile.
(Review ID: 98616)
======================================================================