dlopen vs. libstdc++, need entry in FAQ?

[Christoph Bugel]

[I'm replying to a digest mail, sorry if it appears messy]

> Hey y'all. There seems to be a large number of dlopen-type questions
> regarding libstdc++-v3. There is also a historical pattern of these
> kinds of questions.
> 
> In particular, 
> 
> libstdc++/6062
> libstdc++/5166
> 
> Both of these work, so I don't think I see any bugs. However, it might
> be a good idea to put in some kind of documentation on how to properly
> do this.

Well, on Linux it works (of couse :-), but PR 5166 is about HP-UX, and I'm not
sure that it actually works there, or what I did wrong when I reported it.  It
is probably, as you suggested, a linker-related issue.  When I get back to my
work I'll try, as you suggested, with a more recent (3.1) toolchain, and I'll
use --with-as pointing to gas (like I did when I reported the PR). And I'll
leave out the --enable-threads this time, because the documentation now
officially states that this doesn't currently work on HP-UX.

The -rdynamic you mention might have to do something with this, but on the
platform where the 'bug'  triggers, I couldn't use this switch!
g++: unrecognized option `-rdynamic'. (and yes I know, on Linux it works)
BTW, GNU-binutils on HP-UX doesn't create an ld binary

So, IMO this is still a `bug'. I'll try to post results with 3.1 ASAP.
Thanks!
Christoph