On Lunes 16 Marzo 2009, Harald Hanche-Olsen wrote:
> + Gábor Melis <mega@...>:
> > I wrote a section that may go into BUGS later on what's needed to
> > diagnose these bugs. I guess, your test case is just:
> > (handler-bind ((sb-ext:timeout #'continue))
> > (sb-ext:with-timeout 3
> > (sleep 2)
> > (sleep 2)))
> Indeed, it seems so.
> I built sbcl with ldb support plus QSHOW, QSHOW_SAFE and
> QSHOW_SIGNAL. Now when I run it, it spews messages like these, but I
> suppose that is to be expected?
> Memory fault at: 0x100fcc9c, PC: 0x1006c780
> heap WP violation? fault_addr=100fcc9c, page_index=252
> Memory fault at: 0x10027344, PC: 0x1006c7c0
> heap WP violation? fault_addr=10027344, page_index=39
> Anyway, it seems to work more-or-less as normal otherwise (when you
> can see the normal output through the thickets of debug spew).
Yes, it's normal.
> If I cut down interface.pure.lisp to contain only the code above,
> then run "sh run-tests.sh --break-on-failure foo.pure.lisp", the run
> ends with
> Memory fault at: 0x103d6a74, PC: 0x10cfe1fc
> heap WP violation? fault_addr=103d6a74, page_index=982
> fatal error encountered in SBCL pid 76297:
> deferrable signal 1 blocked
> test failed, expected 104 return code, got 1
> and sbcl quits, so there is nothing there to debug.
Remove the --disable-debugger from tests/subr.sh for now.
> OTOH, if I just run sbcl as
> ./src/runtime/sbcl --core output/sbcl.core --no-sysinit --no-userinit
Yeah, I meant this way of running it but if it doesn't fail ...
> and paste the above code into it, the code seems to work just fine.
> Again, there seems to be nothing to debug. It's a Heisenbug?
> I don't know what to try next.
> - Harald