This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Whopr versus weakref


> On Tue, 20 Apr 2010, Jan Hubicka wrote:
> 
> > Hi,
> > buiding whopr I get error on weakref alias in pthread library.  We must bring
> > functions public for ltrans split yet weakrefs are defined for static variables
> > only.  I am not sure what to do here, but I think it is safe to weaken the
> > check for ltrans so we always end up with weak reference.
> > 
> > Or better ideas?
> 
> Can't we avoid bringing these functions public by putting them into
> a proper partition?  Otherwise we risk breaking code when we suddenly

Those are variables, and, no at the moment we can put them into proper partition
since we have no idea who is referencing them (we make no track about what is taking
address of what that I would like to add).  Even after that this would interfere
badly with partitioning, since it would cause the variable and everything that possible
access it to be forced into a single partition that is relatively bad.

So I would like to see this solved differently.

> change symbol linkage, no?  Think of using whopr to do partial
> linking of {a.c, b.c} -> x.o and {c.c, d.c} -> y.o and then linking
> x.o and y.o.

No idea here, will need to think of what weakref is really about :)
But they are all static, so linking of a.c, b.c->x.o should resolve the weakref
same way as it would if we linked a.c only...

Honza


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]