|
From: Josef W. <Jos...@gm...> - 2005-11-18 20:30:45
|
On Friday 18 November 2005 20:58, Julian Seward wrote:
> I agree. I was just experimenting with this:
>
> if (0==VG_(strcmp)("Memcheck", VG_(details).name)) {
>
> add_redirect_sym_to_addr(
> "soname:ld.so.1", "strlen",
> (Addr)&VG_(ppc32_linux_REDIR_FOR_strlen)
> );
> add_redirect_sym_to_addr(
> "soname:ld.so.1", "strcmp",
> (Addr)&VG_(ppc32_linux_REDIR_FOR_strcmp)
> );
>
> }
>
> Would that be an acceptable temporary fix for 3.1.0 that you'd be
> happy with?
Sure. Thanks.
Thinking about it, redirection needs are really specific to tools.
In the next VG version, they should be installed by the tool instead of the
core, as Nick said.
BTW: I just found a bug regarding the call graph on PPC. I have a problem
with conditional returns. They always are converted to conditional jumps
(which can get calls depending on the situation).
Of course this is because x86/amd64 does not have conditional returns.
If you do not receive a new RC from me by the time you want to do the
announcement, please still use the RC I sent to you yesterday.
I just thought about the BB discarding problem: Currently, I keep counters
for discarded BBs in the hope that the code will be mapped again. This is bad.
A solution would be to simply output the counters of these BBs at discarding
time. This probably is also a nice idea for cachegrind...
Josef
|