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: IPCP & versioning additions and fixes for tree profiling branch


> On Sunday 20 March 2005 17:54, Jan Hubicka wrote:
> > > > What exactly this is shooting for?  Won't we run into unwanted
> > > > collisions or something like that?
> > >
> > > No, because a unique number is appended.  It's ugly though, but
> > > we talked about this last week and couldn't think of anything
> > > better.
> >
> > Hmm, I am bit getting lost here how the wrong characters come in at
> > first place?  Do you use something like unmangled identifier instead of
> > DECL_ASSEMBLER_NAME or so?
> 
> Yes.  The DECL_ASSEMBLER_NAME of the parent is not available at this
> point to name the clone, and my guess is that setting it causes an ICE
> if it later turns out that the parent is not necessary.
> And it is in any case utterly useless to use the mangled name because
> after appending the number you cannot demange it anyway.  See here the
> reason why the ABI extension is needed for versioning ;-)

OK, lets go ahead with this then (I don't see why DECL_ASSEMBLER_NAME
must lead to ICE  - last time I tried it was possible to set it for
anything you wanted and definitly we should not create dead clones), but
I guess we can cope with all this later.

Honza
> 
> Gr.
> Steven
> 
> 


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