[libstdc++] Re: static library for testsuite runs

[Phil Edwards]

I'm applying this patch, originally posted on the v3 list.  More notes:

> This does not use libtool.  Should it?  I have no particular desire to
> take the two most opaque GNU projects (dejagnu and libtool) and figure
> out from first principles how to tie them together.

This question stands.

> Also, this should be tested on a platform which uses multilibs.  I have
> not been able to build on sparc since before I developed the patch, and
> I have no other multilib-using platform available.

I have since hacked my i686-pc-linux-gnu config to be a multilib platform
(that was, ah, educational), and this patch causes no regressions there
and works fine with the multilibs.


> As I mentioned on the list before 3.1, I've been using this for a while now.
> My initial motivation was bringing the testsuite results for concept-checking
> into line with those without concept-checking.  (C-c adds two regressions.
> This fixes one problem in passing; one remains.)

The testing was done under i686-pc-linux-gnu (both normal and hacked up).
It's going on the trunk only, unless we really need it to fix something
else on the branch.


2002-05-28  Phil Edwards  <pme@gcc.gnu.org>

	* testsuite/Makefile.am (noinst_LIBRARIES):  New target.  Pull in
	CXX/INCLUDES.
	* testsuite/Makefile.in:  Regenerate.
	* testsuite/testsuite_hooks.h (gnu_copy_tracker):  Move from
	list_modifiers.cc and rename from 'T'.  Move code bodies...
	* testsuite/testsuite_hooks.cc:  ...to here.  New file.
	* testsuite/23_containers/list_modifiers.cc:  Move 'T' class out.
	* testsuite/lib/libstdc++-v3-dg.exp (libstdc++-v3_target_compile):
	Add libv3test.a to link options.