Re: VIRT_ADDR_VARIES

[Paul Eggert]

On 11/08/11 10:38, Andreas Schwab wrote:
> It causes PURE_P to always return true.

OK, so Emacs is dumping core, not valgrind.
But how can valgrind affect the address of my_edata?
Valgrind supplies its own allocator, but I didn't
think it could move initialized data around.

I'm asking partly because I'm worried that if
valgrind can move my_edata, then there are more-serious
issues that we also would need to address, i.e. that
the PURE_P core dump is just the tip of an iceberg.

I assume this is on x86-64 GNU/Linux?  I'm not observing
this behavior on my host, which is that flavor.
I'm running Fedora 15, which has valgrind 3.6.1.