This is the mail archive of the
mailing list for the libstdc++ project.
Re: Revert __use_cache relocation
Carlo Wood writes:
> On Tue, Jul 01, 2003 at 11:51:47AM +0200, Carlo Wood wrote:
> > On Tue, Jul 01, 2003 at 01:38:53AM -0400, Jerry Quinn wrote:
> > > This patch http://gcc.gnu.org/ml/libstdc++/2003-06/msg00368.html is
> > > causing the instantiation problems I'm getting in
> > > http://gcc.gnu.org/ml/libstdc++/2003-07/msg00003.html
> > >
> > > Reverting this patch until we can figure out how to solve this.
> > >
> > > Index: include/bits/locale_facets.tcc
> > > + const __numpunct_cache<char>&
> > > + __use_cache(const locale& __loc);
> > If you put this in locale_facets.tcc and #include
> > it in a compilation unit, then it will instantiate
> > std::numpunct<char>::id no?
> > So, if you include it in two different compilation
> > units (two different *.cc file #include it), then
> > you get a redefinition.
The library files get built with -fno-implicit-instantiation. The one
exception is concept-inst.cc. That's so that it can instantiate a
whole raft of stuff to perform concept checks (and beyond that, I
don't know what's going on in there).
The funny thing is, moving the code into the header file worked fine a
couple of days ago. That was before the recent flurry of qualifying
everything. I guess I'll have to do a search to find which patch
changes things. Alternatively, concept-inst.cc could be forced to
explicitly instantiate anything that's missing.
Hey, wait... isn't the linker supposed to clean multiple instantiation
> Atually, I didn't even know if they was a normal
> patch or a reversed patch - so I don't know that
> code is added or removed :).
It was reversed, moving the code back into the .cc file :-)