Data corruption problem compiling Ada rts g-awk

[Jeffrey A Law]

On Mon, 2004-09-13 at 16:19, Richard Kenner wrote:
> When I compile it, I get a SIGSEGV in the GC mark routines.  It
> looks like a PHI_NODE has a bb field that points to memory
> that contains all 0x5a5a5a5a.  I tried to set a breakpoint at the
> allocation of that node but it never tripped.
> 
> Has anybody else seen this?
Diego seemed to have tripped over it too.

The symptoms you describe are consistent with an object being
freed while there are still references to the object.

What's the full backtrace?  Does the mark routine for dataflow_d
appear in the backtrace?  If so it could well be the same problem
Diego tripped over.

jeff