GCJ application runs not too bad with shared libs, but crasheswith static libs - seems to be in _Jv_MonitorEnter

[Adam Megacz]

Andrew Haley <aph@redhat.com> writes:
> In any case, here's a question: if we lose line numbers in stack
> traces, is that a disaster?  Does it matter?

I wouldn't recommend making my stuff the default.  I recall porting
XWT to the MSJVM, which doesn't print line numbers, and debugging was
much, much, much more painful.

The way it used to work, my code would only execute if forking
addr2line failed.

Here's a crazy idea: could we insert a NOP into the method body at a
location corresponding to each line break?  Then all we'd need is the
starting line number for each method in order to get stack trace line
numbers.

The NOPs wouldn't slow down execution.  Would they introduce an
unacceptable delay due to increased prefetch size?  Would it
substantially bloat the binary?

  - a