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]

[C++ PATCH] Fix spew.c GC ICE (take 2)


On Tue, Jan 22, 2002 at 07:51:55PM +0000, Jason Merrill wrote:
> >>>>> "Jakub" == Jakub Jelinek <jakub@redhat.com> writes:
> 
> > On Tue, Jan 22, 2002 at 05:50:01PM +0000, Jason Merrill wrote:
> >> >>>>> "Jakub" == Jakub Jelinek <jakub@redhat.com> writes:
> >> 
> >> > 	(do_pending_inlines): Set last for the last pending_inlines in the
> >> > 	chain, chain current processing_these_inlines to it.
> >> 
> >> Would it would make sense to save processing_these_inlines in struct
> >> cp_function?  Maybe not; there doesn't seem to be any precedent for saving
> >> variables that relate to the enclosing context.
> 
> > It is certainly possible, yes. If you prefer that as a cleaner solution, I
> > can code it up.
> 
> I think I would prefer that, thanks.

Below, bootstrap pending, ok to commit if no regressions?

> >> > BTW: is the unparsed_text made by snarf_defarg registered with GC somehow?
> >> > Doesn't look like that to me...
> >> 
> >> It's attached to the parameter it is the default for.
> 
> > But nobody is ever calling mark_pending_inlines on it, so it is not.
> 
> But it's attached to the parameter, which is attached to the function,
> which is attached to the class, which is attached to the namespace, which
> is attached to the global list of decls, which is a GC root.  Right?

Yes, registered with GC is not a good term for what I've been asking.
I wanted to know if somebody ever calls ggc_mark_tree on the trees stored
within, and it looks like nobody is doing this. I'll write a separate patch
for this.

	Jakub

Attachment: P3
Description: Text document


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