Re: [PATCH] Omit frame pointer and fix %ebp by default on x86 (take3)

[Bryce McKinlay <mckinlay at redhat dot com>]

Daniel Berlin wrote:

> > Well, missing some stack frames is, from the Java point of view,
> > failing.  We need stack traces for correct functioning.
> >
> > We will work around this in Java by using unwinder info to do a
> > backtrace, but we don't yet do so.  Please do not check in this patch.
>
>
> When?
> It's now been 6 months since the variable tracking stuff necessary to 
> get omit-frame-pointer + debug info going was committed (not to 
> mention that it was on the rtlopt branch for a while before then).
> Soon after, a patch to turn on omit-frame-pointer by default was 
> submitted and shot down for the same reason you proffer now.
> It would be good to know the timeframe when you plan to have java use 
> unwinder info to do a backtrace, so that people stop submitting this 
> patch (among other obvious reasons).


libgcj will be changed to use unwinder info pretty soon now. All the 
code has been written, there are just a couple of issues to resolve 
before we make the switch. I don't have an exact time frame but I really 
want to get it in as soon as possible.

If you want to make this change now, it should be pretty trivial to 
change the Java front end to always enable -fno-omit-frame-pointer, right?

As for glibc backtrace(), couldn't that be changed to use the unwinder 
info as well?

Regards

Bryce