[Ltrace-devel] [PATCH v2] Tracing PLT-less MIPS binaries

Petr Machata pmachata at redhat.com
Fri Feb 6 00:04:49 UTC 2015


Petr Machata <pmachata at redhat.com> writes:

> Faraz Shahbazker <faraz.shahbazker at imgtec.com> writes:
>
>> 2. mips_unresolved_data is shallow copied for each cloned symbol. If a
>> child detaches whilst a symbol remains in NEED_UNRESOLVE state, then
>> we shouldn't free its unresolve_data because the parent/siblings may 
>> still need it. I've added a ref_count field to the struct to fix this.
>> I am not sure how(if at all) PPC tackles this problem.
>
> Looking in, I don't think it does.  That's probably a bug, I'll need to
> address that.  Good catch!

Uh, of course it does.  I _though_ this sounded familiar--I was fixing a
bug in that cloning recently.  But I was looking at pmachata/mips where
the bug wasn't yet fixed ;)

> The refcounting is fine by me, but for ppc I'll just do deep copy.
> Easier to get right, I feel.

It does a deep copy.

Thanks,
Petr



More information about the Ltrace-devel mailing list