On 2015/11/25 01:41:13, Yang wrote:
On 2015/11/25 00:04:51, caitp wrote:
> On 2015/11/24 21:37:29, caitp wrote:
> > +yang, do you have an idea what would cause the DCHECK failure in
> > VerifyRecompiledCode()? It doesn't seem to break the code when
DCHECKS are
> > disabled.
>
> More context, old_target->kind() == BUILTIN, while new_target->kind() ==
STUB
Caitlin,
this check makes sure that when we recompile code for debugging (to get
code
with debug break slots), the new code is in sync with the old one so that
we
can
find the matching pc when replacing old code while it is on the stack. We
match
up positions in both code objects by counting calls up to the current pc.
Therefore we expect each call to correspond to each other.
If you are not doing weird tricks like patching call targets after
compile,
the
corresponding call (i.e. same number of calls preceding it) should not
target
different code kinds.
Let me know if you found anything. I can try to figure this one out in the
morning if you havent yet.
Btw this likely happens because the recompile took a different code path.
That
could be because of lazy vs eager parsing, because of debugger being
active, or
maybe because of subtle inderminism in the parser/compiler.
https://codereview.chromium.org/1309813007/
--
--
v8-dev mailing list
v8-dev@googlegroups.com
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to v8-dev+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
