This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
Re: g++/cygwin-B19 linkage problem, multiple definitions
- To: Jason Merrill <jason at cygnus dot com>
- Subject: Re: g++/cygwin-B19 linkage problem, multiple definitions
- From: Mumit Khan <khan at xraylith dot wisc dot edu>
- Date: Thu, 29 Oct 1998 16:33:26 -0600
- Cc: Mumit Khan <khan at cygnus dot com>, law at cygnus dot com, Jan Reimers <janr at molienergy dot bc dot ca>, "'egcs-bugs at cygnus dot com'" <egcs-bugs at cygnus dot com>
Jason Merrill <jason@cygnus.com> writes:
> >>>>> Mumit Khan <khan@xraylith.wisc.edu> writes:
>
> > Also, I may be completely off the base here, but wasn't there something
> > about MIPS port not liking this change (and hence Jason's Oct 13 change
> > to mips/iris6.h and varasm.c)?
>
> Yes. And you'll also need
>
> * decl2.c (finish_vtable_vardecl): Check DECL_INTERFACE_KNOWN
> instead of linkage.
>
> I wasn't planning to put this change in because of the follow-on patches it
> required, but I could be convinced if it's a serious problem.
>
Well, I have a folder full of bug reports that all lead to this. However,
it has a trivial workaround (define a destructor, even if empty), so it's
not crucial. The trouble is when you're using a library with such cases,
and then you're stuck.
I did run the testsuite and no regressions on x86-win32 and ix86-linux.
Of course, the testsuite doesn't really test for lots of things that can
go wrong here.
Mumit