[Openmcl-devel] ARM testing
gb at clozure.com
Tue Jan 25 17:29:14 UTC 2011
On Tue, 25 Jan 2011, David Brown wrote:
> Any idea why this doesn't happen with the non-bootstrap image, or have I
> just not pushed it hard enough?
It's likely that the bug has to do with determining stack bounds for
the initial thread. The bootstrapping image runs entirely on that
initial thread, so whenever the GC runs it runs on that thread.
In the full image, the initial thread creates a listener thread and then
spends most of its time asleep. (It wakes up a few times each second
to force output to standard interactive streams, to check for ^C, and
do other housekeeping tasks.) Those things might cons, trigger a GC,
and run into the same bug, but it's much more likely that a GC in the
full image will run on the listener thread.
More information about the Openmcl-devel