[Gc] Re: GC issue in mono 0.31

Nikolai Zhubr s001 at hotbox.ru
Thu May 6 03:52:42 PDT 2004


Hans,

I've checked __data_start and _end and they appear to be
exactly equal to the values passed to GC_add_roots_inner.

Regarding 6.3alpha6 version, yes it looks like gctest now
passes successfully on 2.6.5 kernel here.
-- 
Best regards,
 Nikolai Zhubr

Thursday, 06 May, 2004, 4:14:38, you wrote:

> Thanks.

> The address range that's passed to GC_add_roots_inner is clearly much too large.

> Could you also print __data_start and _end at the call to GC_add_roots_inner?
> (They're both declared as arrays.  I want the address, not the elements.)
> If that's indeed where the two arguments are coming from, then we fairly
> clearly have a linker issue.

> I put gc6.3alpha6 on the web site.  Hopefuly the other 2.6 problem should
> disappear with that.

> Hans

> On Thu, 6 May 2004, Nikolai Zhubr wrote:

>> Hello Hans,
>> You are right probably that there are actually two separate
>> problems (I was lucky enough to hit both).
>> Returning to dynamic linking issue, attached is the output
>> with GC_add_roots_inner arguments included + /proc/<mono>/maps.
>> (Some addresses have probably changed somewhat, so it might not
>> exactly match previously posted nm output)
>> 




More information about the Gc mailing list