Re: Fix libstdc++ testing using upstream dejagnu

[Hans-Peter Nilsson]

On Wed, 19 May 2004, Hans-Peter Nilsson wrote:
> On Wed, 19 May 2004, Daniel Jacobowitz wrote:
> > http://mail.gnu.org/archive/html/bug-dejagnu/2002-10/msg00009.html

> I'll remove the patch for a known stable tree such as gcc-3.3
> branch (as opposed to current gcc HEAD) and see if there's still
> fallout.

The patch mentioned above, minus the changes to proc
g++_include_flags and proc libstdc++_include_flags still works
for me; no regressions on my gcc-3.3 checkout, for simulator
toolchains for MMIX, MIPS, CRIS.  The CRIS case had newlib and
libgloss installed (at --prefix=), checking that the change
would be harmless for that case.

So your suggestion to only -isystem the libgloss and newlib
headers seems right.

Having said that, I think it's likely that -isystem for the
libstdc++ headers may change and be necessary in the future; I
don't see a conceptual difference to the C/libgloss/newlib case.
(Still, let's not burn that bridge until we're there.  8-)

brgds, H-P