harmony-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Ilya Berezhniuk" <ilya.berezhn...@gmail.com>
Subject Re: [drlvm][signals] New heuristics for crash handler
Date Tue, 12 Feb 2008 21:32:11 GMT
Tim,

As an author of the patch, I confirm that these changes affect the
code working in case of crash.

The only change affecting general VM code is the following one-line change:
-    {OpcodeInfo::em64t,   {REX_W, 0xFF, _2}, {r_m64},     U },
+    {OpcodeInfo::em64t,   {0xFF, _2},        {r_m64},     U },

Both Intel and AMD specs say that REX prefix is not required here - so
I hope this change is harmless too.

Thanks,
Ilya.


2008/2/12, Tim Ellison <t.p.ellison@gmail.com>:
> Gregory Shimansky wrote:
> > Tim Ellison said the following on 12.02.2008 22:32:
> >> Gregory,
> >>
> >> Is this appropriate during feature freeze?  Looks like significant new
> >> functionality.
> >
> > Oh. I didn't realize the _feature_ freeze started up already! I only
> > remembered the date 15th of February from your email. I must have not
> > read it carefully enough. The good thing is that this is not a _feature_
> > per say since crash handler had problems printing good stack traces on
> > x86_64, so it could be called a bug to some extent. On the other hand,
> > if I was aware about the freeze, I'd hold on committing such big change.
> >
> >> Do you guarantee all this code going to work perfectly ?! <g>
> >
> > Well, the code in question is targeted to work only in case of a crash
> > which shouldn't happen in the first place. So I hope it is not critical
> > for general stability. If any regression is noticed, I'll revert this
> > commit.
>
> np -- at least you know there is a feature freeze on now.  Sorry for not
> making it clear.
>
> Regards,
> Tim
>

Mime
View raw message