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: libstdc++ bootstrap failure, today's 3.2 on sparcv9-sun-solaris2.8


On May 24,  2:00am, Richard Henderson wrote:

> On Fri, May 24, 2002 at 08:33:36PM +1200, Bryce McKinlay wrote:
> > This is a problem with dbxout, it doesn't happen using dwarf2. It seems 
> > that dbxout doesn't handle a typedef who's DECL_CONTEXT is a namespace. 
> > This patch makes everything build. Note that dbxout is one of only two 
> > places in the compiler where decl_type_context() is used - the other is 
> > in cp/typeck.c. OK to commit?
> > 
> > 2002-05-24  Bryce McKinlay  <bryce@waitaki.otago.ac.nz>
> > 
> >     * tree.c (decl_type_context): Return NULL_TREE if decl's context is a
> >     namespace.
> 
> Yes, this is ok to get back to bootstrapville.
> 
> Kevin, was this something you were intending to handle?

It wasn't my intent to handle namespaces so Bryce's patch is okay with
me.

That said, I think that decl_type_context() ought to be doing
something other than bailing out by returning NULL_TREE for
namespaces.  I.e, it should either stop and return the namespace's
context or it should follow it (in the loop) as is done for TYPE_DECL
and BLOCK, but I don't know which.  (If it did the former,
dbxout_class_name_qualifiers() could be extended to work for
namespaces too.)

Kevin


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